fix: resolve mapper configuration for ToolDefinitionManifest and Workspace

Move ToolDefinitionManifest import out of TYPE_CHECKING in tool_type.py
so SQLAlchemy can resolve the string-annotated relationship during mapper
configuration.

Add Workspace to models/__init__.py (before ToolInstance) so the
ToolInstance-Workspace relationship can be resolved.

Quality gates: py_compile passed, ruff passed, all mappers configure OK.

.atl/.skill-registry.cache.json

@@ -1,3 +1,3 @@
 {
-  "fingerprint": "fdea8a74bb4c7449c01c4bd61646c895b10ede78"
+  "fingerprint": "c36b11ec5edebc02aa51b1113a7a11dc2559e812"
 }

.atl/skill-registry.md

@@ -2,7 +2,7 @@
 
 <!-- Auto-generated by gentle-pi extensions/skill-registry.ts. Run /skill-registry:refresh to regenerate. -->
 
-Last updated: 2026-05-28
+Last updated: 2026-06-02
 
 ## Sources scanned
 
@@ -21,7 +21,6 @@ Last updated: 2026-05-28
 | Skill | Trigger / description | Scope | Path |
 | --- | --- | --- | --- |
 | `auto-commit` | Use when you are making multiple edits or completing significant work in a git repository to automatically create commits | user | `/home/alex/.config/opencode/skills/auto-commit/SKILL.md` |
-| `openspec` | Use OpenSpec as the source of truth for planning, implementation, verification, and archive discipline. | user | `/home/alex/.config/opencode/skills/openspec/SKILL.md` |
 | `openspec-apply-change` | Implement tasks from an OpenSpec change. Use when the user wants to start implementing, continue implementation, or work through tasks. | project | `/home/alex/projects/headquarter/.opencode/skills/openspec-apply-change/SKILL.md` |
 | `openspec-archive-change` | Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete. | project | `/home/alex/projects/headquarter/.opencode/skills/openspec-archive-change/SKILL.md` |
 | `openspec-explore` | Enter explore mode - a thinking partner for exploring ideas, investigating problems, and clarifying requirements. Use when the user wants to think through something before or during a change. | project | `/home/alex/projects/headquarter/.opencode/skills/openspec-explore/SKILL.md` |

AGENTS.md

@@ -75,6 +75,7 @@ Do not:
 * Introduce new dependencies without clear justification.
 * Treat existing code as more authoritative than OpenSpec for intended behavior.
 * Decide product behavior silently when the spec is unclear.
+* Run `docker compose` commands (build, up, down, etc.) without explicit user approval and proper isolation (e.g., feature branches, separate worktrees, or staged rollouts). Docker Compose operations are deployment-level changes that can affect running services, shared volumes, and network state. Always ask first.
 
 If scope must change, propose an OpenSpec update first.
 

apps/api/Dockerfile

@@ -50,8 +50,8 @@ ENV PATH=/root/.local/bin:$PATH
 # Copy application code
 COPY --chown=appuser:appgroup . .
 
-# Create directories for repo and instance storage
-RUN mkdir -p /data/repos /data/instances && chown -R appuser:appgroup /data
+# Create directories for repo, instance, and workspace storage
+RUN mkdir -p /data/repos /data/instances /data/working-copies && chown -R appuser:appgroup /data
 
 # Copy wait-for-db script
 COPY wait-for-db.sh /usr/local/bin/wait-for-db.sh

apps/api/src/api/events.py

@@ -16,7 +16,7 @@ router = APIRouter(prefix="/events", tags=["events"])
 
 # In-memory connection counter per user (single-process assumption)
 _connection_counts: dict[uuid.UUID, int] = {}
-MAX_CONNECTIONS_PER_USER = 5
+MAX_CONNECTIONS_PER_USER = 20
 
 
 @router.get("/stream")

apps/api/src/api/git_repositories.py

@@ -10,7 +10,12 @@ from pydantic import BaseModel, ConfigDict
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from src.auth.dependencies import _get_owned_project, _get_user, get_current_user_id, get_db_session
+from src.auth.dependencies import (
+    _get_owned_project,
+    _get_user,
+    get_current_user_id,
+    get_db_session,
+)
 from src.config import Settings
 from src.models.git_repository import GitRepository
 from src.models.ssh_key import SSHKey
@@ -90,7 +95,9 @@ def _prepare_ssh_env(ssh_key: SSHKey | None) -> dict | None:
     return env, key_path
 
 
-def _preflight_remote_repository(remote_url: str, ssh_key: SSHKey | None = None) -> None:
+def _preflight_remote_repository(
+    remote_url: str, ssh_key: SSHKey | None = None
+) -> None:
     """Verify a remote repository is reachable before cloning."""
     env = None
     key_path = None
@@ -109,22 +116,32 @@ def _preflight_remote_repository(remote_url: str, ssh_key: SSHKey | None = None)
             env={**os.environ, **env} if env else None,
         )
     except subprocess.TimeoutExpired:
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="remote repository check timed out")
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail="remote repository check timed out",
+        )
     except FileNotFoundError:
-        raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail="git command not found")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="git command not found",
+        )
     finally:
         if key_path and os.path.exists(key_path):
             os.unlink(key_path)
 
     if result.returncode != 0:
-        logger.error("Preflight check failed for %s: stderr=%s", remote_url, result.stderr)
+        logger.error(
+            "Preflight check failed for %s: stderr=%s", remote_url, result.stderr
+        )
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=f"repository not found or inaccessible: {result.stderr}",
         )
 
 
-def _clone_working_repository(remote_url: str, repo_path: str, ssh_key: SSHKey | None = None) -> None:
+def _clone_working_repository(
+    remote_url: str, repo_path: str, ssh_key: SSHKey | None = None
+) -> None:
     env = None
     key_path = None
 
@@ -142,9 +159,14 @@ def _clone_working_repository(remote_url: str, repo_path: str, ssh_key: SSHKey |
             env={**os.environ, **env} if env else None,
         )
     except subprocess.TimeoutExpired:
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="clone operation timed out")
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST, detail="clone operation timed out"
+        )
     except FileNotFoundError:
-        raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail="git command not found")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="git command not found",
+        )
     finally:
         if key_path and os.path.exists(key_path):
             os.unlink(key_path)
@@ -165,7 +187,10 @@ def _init_working_repository(repo_path: str) -> None:
             text=True,
         )
     except FileNotFoundError:
-        raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail="git command not found")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="git command not found",
+        )
 
     if result.returncode == 0:
         return
@@ -310,7 +335,10 @@ async def create_external_repository(
         )
     )
     if existing.scalar_one_or_none():
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="repository name already exists")
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail="repository name already exists",
+        )
 
     # Validate and potentially correct the URL
     remote_url = data.remote_url
@@ -336,13 +364,21 @@ async def create_external_repository(
         try:
             ssh_key_id = uuid.UUID(data.ssh_key_id)
         except ValueError:
-            raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="invalid ssh_key_id format")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail="invalid ssh_key_id format",
+            )
 
         ssh_key = await session.get(SSHKey, ssh_key_id)
         if ssh_key is None:
-            raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found")
+            raise HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found"
+            )
         if ssh_key.user_id != user_id:
-            raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail="ssh key does not belong to user")
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="ssh key does not belong to user",
+            )
 
     if remote_url:
         _preflight_remote_repository(remote_url, ssh_key)
@@ -369,7 +405,10 @@ async def create_external_repository(
             repo.is_mirror = False
         except Exception as exc:
             await session.rollback()
-            raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=f"Failed to clone repository: {exc}")
+            raise HTTPException(
+                status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+                detail=f"Failed to clone repository: {exc}",
+            )
     else:
         # Initialize empty repo
         os.makedirs(repo_path, exist_ok=True)
@@ -438,7 +477,9 @@ async def delete_repository(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     # Remove from disk
     if os.path.exists(repo.path):
@@ -484,7 +525,10 @@ async def create_repository(
         )
     )
     if existing.scalar_one_or_none():
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="repository name already exists")
+        raise HTTPException(
+            status_code=status.HTTP_400_BAD_REQUEST,
+            detail="repository name already exists",
+        )
 
     # Validate and potentially correct the URL
     remote_url = data.remote_url
@@ -511,13 +555,21 @@ async def create_repository(
         try:
             ssh_key_id = uuid.UUID(data.ssh_key_id)
         except ValueError:
-            raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="invalid ssh_key_id format")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail="invalid ssh_key_id format",
+            )
 
         ssh_key = await session.get(SSHKey, ssh_key_id)
         if ssh_key is None:
-            raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found")
+            raise HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found"
+            )
         if ssh_key.user_id != user_id and ssh_key.project_id != project_id:
-            raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail="ssh key does not belong to user or project")
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="ssh key does not belong to user or project",
+            )
 
     if remote_url:
         _preflight_remote_repository(remote_url, ssh_key)
@@ -581,20 +633,30 @@ async def update_repository_ssh_key(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     # Validate SSH key if provided
     if data.ssh_key_id:
         try:
             ssh_key_id = uuid.UUID(data.ssh_key_id)
         except ValueError:
-            raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail="invalid ssh_key_id format")
+            raise HTTPException(
+                status_code=status.HTTP_400_BAD_REQUEST,
+                detail="invalid ssh_key_id format",
+            )
 
         ssh_key = await session.get(SSHKey, ssh_key_id)
         if ssh_key is None:
-            raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found")
+            raise HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND, detail="ssh key not found"
+            )
         if ssh_key.user_id != user_id and ssh_key.project_id != project_id:
-            raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail="ssh key does not belong to user or project")
+            raise HTTPException(
+                status_code=status.HTTP_403_FORBIDDEN,
+                detail="ssh key does not belong to user or project",
+            )
 
         repo.ssh_key_id = ssh_key_id
     else:
@@ -640,16 +702,24 @@ async def get_repository_history(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
-        history = get_commit_history(repo.path, branch=branch, limit=limit, offset=offset)
+        history = get_commit_history(
+            repo.path, branch=branch, limit=limit, offset=offset
+        )
         return history
     except RuntimeError as e:
-        raise HTTPException(status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e))
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e)
+        )
 
 
 @router.get(
@@ -681,10 +751,14 @@ async def get_repository_commit(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         detail = get_commit_detail(repo.path, commit_hash)
@@ -763,10 +837,14 @@ async def list_repository_files(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         entries = list_tree(repo.path, branch=branch, path=path)
@@ -829,10 +907,14 @@ async def get_repository_file_content(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         file_content = get_file_content(repo.path, branch=branch, path=path)
@@ -847,7 +929,9 @@ async def get_repository_file_content(
             last_commit=file_content.last_commit,
         )
     except FileNotFoundError:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="file not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="file not found"
+        )
     except RuntimeError as e:
         raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(e))
 
@@ -880,32 +964,104 @@ async def get_repository_branches(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
-
-    if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
-
-    try:
-        branches, default_branch = list_branches(repo.path)
-        return BranchesResponse(
-            branches=[
-                {
-                    "name": b.name,
-                    "is_default": b.is_default,
-                    "last_commit": b.last_commit,
-                }
-                for b in branches
-            ],
-            default_branch=default_branch,
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
         )
-    except RuntimeError as e:
-        logger.error(
-            "Failed to list branches for repo %s: %s",
-            repo_id,
-            str(e),
-            exc_info=True,
-        )
-        raise HTTPException(status_code=status.HTTP_400_BAD_REQUEST, detail=str(e))
+
+    # Try local repo first (.git subdir for normal repos, HEAD for bare)
+    is_valid_git_repo = os.path.isdir(
+        os.path.join(repo.path, ".git")
+    ) or os.path.isfile(os.path.join(repo.path, "HEAD"))
+
+    if is_valid_git_repo:
+        try:
+            branches, default_branch = list_branches(repo.path)
+            return BranchesResponse(
+                branches=[
+                    {
+                        "name": b.name,
+                        "is_default": b.is_default,
+                        "last_commit": b.last_commit,
+                    }
+                    for b in branches
+                ],
+                default_branch=default_branch,
+            )
+        except RuntimeError as e:
+            logger.error(
+                "Failed to list branches for repo %s: %s",
+                repo_id,
+                str(e),
+                exc_info=True,
+            )
+            raise HTTPException(
+                status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=str(e)
+            ) from e
+
+    # Local repo missing/corrupt — try remote if available
+    if repo.remote_url:
+        ssh_key = None
+        if repo.ssh_key_id:
+            ssh_key = await session.get(SSHKey, repo.ssh_key_id)
+
+        ssh_result = _prepare_ssh_env(ssh_key)
+        env = None
+        key_path = None
+        if ssh_result:
+            env, key_path = ssh_result
+
+        try:
+            result = subprocess.run(
+                ["git", "ls-remote", "--heads", repo.remote_url],
+                capture_output=True,
+                text=True,
+                timeout=30,
+                env={**os.environ, **env} if env else None,
+            )
+            if result.returncode == 0:
+                remote_branches = []
+                default_branch = "main"
+                for line in result.stdout.strip().split("\n"):
+                    if line:
+                        parts = line.split("\t")
+                        if len(parts) == 2:
+                            ref = parts[1]
+                            if ref.startswith("refs/heads/"):
+                                branch_name = ref[len("refs/heads/") :]
+                                remote_branches.append(branch_name)
+                                if branch_name in ("main", "master"):
+                                    default_branch = branch_name
+                if remote_branches:
+                    return BranchesResponse(
+                        branches=[
+                            {
+                                "name": b,
+                                "is_default": b == default_branch,
+                                "last_commit": None,
+                            }
+                            for b in remote_branches
+                        ],
+                        default_branch=default_branch,
+                    )
+            else:
+                logger.warning(
+                    "ls-remote returned %d for repo %s: %s",
+                    result.returncode,
+                    repo_id,
+                    result.stderr,
+                )
+        except subprocess.TimeoutExpired:
+            logger.warning("ls-remote timed out for repo %s", repo_id)
+        except Exception as e:
+            logger.warning("ls-remote failed for repo %s: %s", repo_id, str(e))
+        finally:
+            if key_path and os.path.exists(key_path):
+                os.unlink(key_path)
+
+    raise HTTPException(
+        status_code=status.HTTP_404_NOT_FOUND,
+        detail="repository not found on disk — re-clone or re-create the repository",
+    )
 
 
 @router.post(
@@ -938,10 +1094,14 @@ async def update_repository_file(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     # Get user info for commit
     user = await _get_user(session, user_id)
@@ -1009,10 +1169,14 @@ async def get_repository_status(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         status_result = get_status(repo.path)
@@ -1068,10 +1232,14 @@ async def create_repository_branch(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         create_branch(repo.path, data.name, data.base_branch)
@@ -1111,10 +1279,14 @@ async def delete_repository_branch(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         delete_branch(repo.path, branch_name, force)
@@ -1152,10 +1324,14 @@ async def checkout_repository_branch(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         checkout_branch(repo.path, data.branch)
@@ -1204,10 +1380,14 @@ async def commit_repository_changes(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     # Get user info for commit
     user = await _get_user(session, user_id)
@@ -1262,10 +1442,14 @@ async def fetch_repository(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         fetch(repo.path)
@@ -1308,10 +1492,14 @@ async def pull_repository(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         pull(repo.path, branch)
@@ -1354,10 +1542,14 @@ async def push_repository(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         push(repo.path, branch)
@@ -1407,10 +1599,14 @@ async def merge_repository_branches(
 
     repo = await session.get(GitRepository, repo_id)
     if repo is None or repo.project_id != project_id:
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found"
+        )
 
     if not os.path.exists(repo.path):
-        raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk")
+        raise HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail="repository not found on disk"
+        )
 
     try:
         commit_hash = merge(

apps/api/src/api/terminal.py

@@ -222,8 +222,7 @@ async def _handle_terminal_websocket(
         # Use mutable session reference so loops can survive reset
         session_ref = SessionRef(session, slot_session_id)
 
-        # Start I/O loops and heartbeat
-        read_task = asyncio.create_task(_read_loop(session_ref, websocket))
+        # Start write loop and heartbeat (read is now event-driven in TerminalSession)
         write_task = asyncio.create_task(
             _write_loop(session_ref, websocket, instance_id)
         )
@@ -232,7 +231,7 @@ async def _handle_terminal_websocket(
 
         # Wait for either task to complete (indicating disconnect or error)
         done, pending = await asyncio.wait(
-            [read_task, write_task, heartbeat_task],
+            [write_task, heartbeat_task],
             return_when=asyncio.FIRST_COMPLETED,
         )
 
@@ -267,28 +266,6 @@ async def _handle_terminal_websocket(
                 )
 
 
-async def _read_loop(session_ref: SessionRef, websocket) -> None:
-    """Read output from the container and send to WebSocket."""
-    try:
-        while True:
-            session = session_ref.session
-            if not session.is_alive() or session._closed:
-                await asyncio.sleep(0.1)
-                continue
-            data = await session.read_output()
-            if data:
-                try:
-                    await websocket.send_bytes(data)
-                except WebSocketDisconnect:
-                    break
-                except Exception:
-                    break
-            else:
-                await asyncio.sleep(0.01)
-    except Exception:
-        pass
-
-
 async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> None:
     """Read input from WebSocket and send to container."""
     try:
@@ -319,6 +296,10 @@ async def _write_loop(session_ref: SessionRef, websocket, instance_id: str) -> N
                                     rows,
                                 )
                                 await session.resize(cols, rows)
+                            elif msg_type == "ack":
+                                char_count = ctrl.get("chars", 0)
+                                if char_count > 0:
+                                    session.acknowledge_data(char_count)
                             elif msg_type == "reset":
                                 # Reset terminal session (scoped to current slot)
                                 logger.debug(

apps/api/src/api/tool_instances.py

@@ -1039,7 +1039,7 @@ services:
             write_compose_file(instance_dir, compose_content)
 
         elif tool_type.definition_type == "manifest":
-            # Manifest-based: build image and generate compose
+            # Manifest-based: generate compose only; image built lazily on start
             from src.models.tool_definition_manifest import ToolDefinitionManifest
 
             manifest_def = await session.get(
@@ -1061,44 +1061,8 @@ services:
                         deep_merge(dict(base_def.manifest), manifest)
                     )
 
-            # Determine home directory for path expansion
-            _home_dir = get_manifest_home_dir(manifest)
-
             image_tag = compute_image_tag(tool_type.name, manifest)
 
-            # Build image during creation so start is fast
-            dockerfile = compile_dockerfile(manifest)
-            entrypoint = compile_entrypoint(manifest)
-            build_ctx = {
-                "Dockerfile": dockerfile,
-                ".headquarter/entrypoint.sh": entrypoint,
-            }
-
-            returncode, stdout, stderr = await asyncio.to_thread(
-                build_image,
-                instance_dir=instance_dir,
-                dockerfile=dockerfile,
-                tag=image_tag,
-                build_context=build_ctx,
-            )
-
-            if returncode != 0:
-                logger.error(
-                    "Failed to build image for manifest instance %s: %s",
-                    instance_name,
-                    stderr,
-                )
-                raise HTTPException(
-                    status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
-                    detail=f"Failed to build Docker image: {stderr[:500]}",
-                )
-
-            logger.info(
-                "Built manifest image %s for instance %s",
-                image_tag,
-                instance_name,
-            )
-
             variables = {
                 "IMAGE_TAG": image_tag,
                 "INSTANCE_NAME": instance_name.lower(),
@@ -1466,6 +1430,18 @@ async def _prepare_manifest_instance(
 
     compose_content = compile_compose(manifest, variables)
 
+    logger.debug(
+        "_prepare_manifest_instance for %s: repo_path=%s compose_volumes=%s",
+        instance.id,
+        repo_path or "<empty>",
+        manifest.get("mounts", []),
+    )
+    logger.debug(
+        "Generated compose for %s:\n%s",
+        instance.id,
+        compose_content,
+    )
+
     # Cache
     instance.image_tag = image_tag
     instance.manifest_compiled_at = datetime.now()
@@ -1635,38 +1611,14 @@ async def start_instance(
 
     # Mount selected SSH keys into container home dir
     if instance.ssh_key_ids:
+        from src.services.ssh_keys import write_ssh_config, _sanitize_filename
+
+        # Collect all valid keys first
+        ssh_keys_to_mount = []
         for key_id in instance.ssh_key_ids:
             ssh_key = await session.get(SSHKey, uuid.UUID(key_id))
             if ssh_key and ssh_key.user_id == user_id:
-                try:
-                    ssh_dir = prepare_ssh_key_files(
-                        instance_dir,
-                        ssh_key,
-                        subdir=f"mounts/ssh/{key_id}/.ssh",
-                        uid=container_uid,
-                        gid=container_gid,
-                    )
-                    ssh_target = os.path.join(home_dir, ".ssh")
-                    extra_volumes.append(
-                        {
-                            "source": ssh_dir,
-                            "target": ssh_target,
-                            "type": "bind",
-                        }
-                    )
-                    logger.debug(
-                        "Mounted SSH key %s for instance %s to %s",
-                        ssh_key.name,
-                        instance.id,
-                        ssh_target,
-                    )
-                except Exception as exc:
-                    logger.error(
-                        "Failed to prepare SSH key %s for instance %s: %s",
-                        key_id,
-                        instance.id,
-                        exc,
-                    )
+                ssh_keys_to_mount.append(ssh_key)
             else:
                 logger.warning(
                     "SSH key %s not found or not authorized for user %s",
@@ -1674,6 +1626,79 @@ async def start_instance(
                     user_id,
                 )
 
+        if ssh_keys_to_mount:
+            # Use a single shared .ssh directory so all keys are visible
+            ssh_dir = os.path.join(instance_dir, "mounts", "ssh", ".ssh")
+            os.makedirs(ssh_dir, exist_ok=True)
+
+            key_filenames = []
+            for ssh_key in ssh_keys_to_mount:
+                # Use sanitized key name as filename prefix to avoid collisions
+                key_name = _sanitize_filename(ssh_key.name)
+                # If multiple keys have the same name, append a short hash
+                base_filename = f"id_ed25519_{key_name}"
+                filename = base_filename
+                counter = 1
+                while filename in key_filenames:
+                    filename = f"{base_filename}_{counter}"
+                    counter += 1
+                key_filenames.append(filename)
+
+                try:
+                    prepare_ssh_key_files(
+                        instance_dir,
+                        ssh_key,
+                        subdir="mounts/ssh/.ssh",
+                        uid=container_uid,
+                        gid=container_gid,
+                        key_filename=filename,
+                        write_config=False,
+                    )
+                    logger.debug(
+                        "Prepared SSH key %s as %s for instance %s",
+                        ssh_key.name,
+                        filename,
+                        instance.id,
+                    )
+                except Exception as exc:
+                    logger.error(
+                        "Failed to prepare SSH key %s for instance %s: %s",
+                        ssh_key.id,
+                        instance.id,
+                        exc,
+                    )
+
+            # Write combined SSH config with all keys
+            try:
+                write_ssh_config(
+                    ssh_dir,
+                    key_filenames,
+                    uid=container_uid,
+                    gid=container_gid,
+                )
+            except Exception as exc:
+                logger.error(
+                    "Failed to write SSH config for instance %s: %s",
+                    instance.id,
+                    exc,
+                )
+
+            # Mount the single .ssh directory into container home
+            ssh_target = os.path.join(home_dir, ".ssh")
+            extra_volumes.append(
+                {
+                    "source": ssh_dir,
+                    "target": ssh_target,
+                    "type": "bind",
+                }
+            )
+            logger.debug(
+                "Mounted %d SSH key(s) for instance %s to %s",
+                len(ssh_keys_to_mount),
+                instance.id,
+                ssh_target,
+            )
+
     # ── MANIFEST-BASED FLOW ──────────────────────────────────────
     resolved_manifest = None
 

apps/api/src/api/workspaces.py

@@ -38,6 +38,7 @@ async def list_all_workspaces(
             Workspace,
             GitRepository.name.label("repo_name"),
             GitRepository.project_id,
+            GitRepository.ssh_key_id.label("repo_ssh_key_id"),
             instance_count.label("instance_count"),
         )
         .join(GitRepository, Workspace.repo_id == GitRepository.id)
@@ -52,6 +53,7 @@ async def list_all_workspaces(
             "name": ws.name,
             "repo_id": str(ws.repo_id),
             "repo_name": repo_name or "",
+            "repo_ssh_key_id": str(ssh_key_id) if ssh_key_id else None,
             "project_id": str(project_id) if project_id else "",
             "project_name": "",
             "user_id": str(ws.user_id),
@@ -63,10 +65,45 @@ async def list_all_workspaces(
             "updated_at": ws.updated_at.isoformat() if ws.updated_at else None,
             "instance_count": count or 0,
         }
-        for ws, repo_name, project_id, count in rows
+        for ws, repo_name, project_id, ssh_key_id, count in rows
     ]
 
 
+@all_workspaces_router.delete("/{workspace_id}")
+async def delete_workspace_top_level(
+    workspace_id: uuid.UUID,
+    force: bool = Query(False),
+    user_id: uuid.UUID = Depends(get_current_user_id),
+    session: AsyncSession = Depends(get_db_session),
+) -> dict:
+    """Delete a workspace via top-level path."""
+    workspace = await session.get(Workspace, workspace_id)
+    if not workspace or workspace.user_id != user_id:
+        raise HTTPException(status_code=404, detail="Workspace not found")
+
+    manager = WorkspaceManager()
+    try:
+        await manager.delete(workspace, force=force, session=session)
+        await session.commit()
+    except WorkspaceHasInstancesError as exc:
+        await session.rollback()
+        raise HTTPException(
+            status_code=409,
+            detail={
+                "message": "Workspace has running tool instances",
+                "instances": exc.instances,
+            },
+        ) from exc
+    except Exception as exc:
+        await session.rollback()
+        logger.error("Failed to delete workspace: %s", exc)
+        raise HTTPException(
+            status_code=500, detail="Failed to delete workspace"
+        ) from exc
+
+    return {"status": "deleted"}
+
+
 @all_workspaces_router.post("/")
 async def create_workspace_top_level(
     data: dict,
@@ -95,7 +132,7 @@ async def create_workspace_top_level(
 
     manager = WorkspaceManager()
     try:
-        workspace = await manager.create(repo, user_id, name, branch)
+        workspace = await manager.create(repo, user_id, name, branch, session=session)
         session.add(workspace)
         await session.commit()
     except Exception as exc:
@@ -155,6 +192,7 @@ async def list_workspaces(
             "name": ws.name,
             "repo_id": str(ws.repo_id),
             "repo_name": repo.name,
+            "repo_ssh_key_id": str(repo.ssh_key_id) if repo.ssh_key_id else None,
             "project_id": str(repo.project_id) if repo.project_id else "",
             "project_name": repo.project.name if repo.project else "",
             "user_id": str(ws.user_id),
@@ -191,9 +229,15 @@ async def create_workspace(
 
     manager = WorkspaceManager()
     try:
-        workspace = await manager.create(repo, user_id, name, branch)
+        workspace = await manager.create(repo, user_id, name, branch, session=session)
         session.add(workspace)
         await session.commit()
+    except HTTPException:
+        raise
+    except ValueError as exc:
+        await session.rollback()
+        logger.error("Failed to create workspace: %s", exc)
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
     except Exception as exc:
         await session.rollback()
         logger.error("Failed to create workspace: %s", exc)
@@ -320,7 +364,7 @@ async def delete_workspace(
             status_code=409,
             detail={
                 "message": "Workspace has running tool instances",
-                "instances": [{"id": str(i.id), "name": i.name} for i in exc.instances],
+                "instances": exc.instances,
             },
         ) from exc
     except Exception as exc:
@@ -346,7 +390,7 @@ async def sync_workspace(
     workspace = await _get_workspace(session, workspace_id, repo_id)
 
     manager = WorkspaceManager()
-    result = await manager.sync(workspace)
+    result = await manager.sync(workspace, session=session)
 
     if result.branch_deleted:
         raise HTTPException(

apps/api/src/models/__init__.py

@@ -12,6 +12,7 @@ from src.models.tool_instance import ToolInstance
 from src.models.tool_type import ToolType
 from src.models.user import User
 from src.models.user_config import UserConfig
+from src.models.workspace import Workspace
 
 __all__ = [
     "Base",
@@ -29,4 +30,5 @@ __all__ = [
     "ToolType",
     "User",
     "UserConfig",
+    "Workspace",
 ]

apps/api/src/models/tool_type.py

@@ -7,8 +7,9 @@ from sqlalchemy.orm import Mapped, mapped_column, relationship
 
 from src.models.base import Base, TimestampMixin, UUIDPrimaryKeyMixin
 
+from src.models.tool_definition_manifest import ToolDefinitionManifest
+
 if TYPE_CHECKING:
-    from src.models.tool_definition_manifest import ToolDefinitionManifest
     from src.models.user import User
 
 

apps/api/src/services/git_service.py

@@ -2,7 +2,9 @@
 
 import asyncio
 import logging
+import os
 import subprocess
+import tempfile
 
 logger = logging.getLogger(__name__)
 
@@ -11,13 +13,39 @@ class GitService:
     """Low-level git operations for creating and syncing workspaces."""
 
     @staticmethod
-    async def clone(remote_url: str, branch: str, path: str) -> None:
+    def _prepare_ssh_env(
+        ssh_key: str | None,
+    ) -> tuple[dict[str, str] | None, str | None]:
+        """Prepare environment for git commands with SSH authentication.
+
+        Returns a tuple of (env_dict, temp_key_path). Caller must clean up key_path.
+        """
+        if not ssh_key:
+            return None, None
+
+        fd, key_path = tempfile.mkstemp(prefix="ssh_key_")
+        try:
+            os.write(fd, ssh_key.encode())
+        finally:
+            os.close(fd)
+        os.chmod(key_path, 0o600)
+
+        env = {
+            "GIT_SSH_COMMAND": f"ssh -i {key_path} -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null"
+        }
+        return env, key_path
+
+    @staticmethod
+    async def clone(
+        remote_url: str, branch: str, path: str, ssh_key: str | None = None
+    ) -> None:
         """Clone a repository to the given path.
 
         Args:
             remote_url: The git remote URL.
             branch: The branch to clone.
             path: The destination path for the clone.
+            ssh_key: Optional decrypted SSH private key for authentication.
 
         Raises:
             RuntimeError: If the clone fails.
@@ -31,88 +59,118 @@ class GitService:
             remote_url,
             path,
         ]
-        proc = await asyncio.create_subprocess_exec(
-            *cmd,
-            stdout=asyncio.subprocess.PIPE,
-            stderr=asyncio.subprocess.PIPE,
-        )
-        stdout, stderr = await proc.communicate()
-        if proc.returncode != 0:
-            error_msg = stderr.decode().strip() if stderr else "unknown error"
-            logger.error("Git clone failed: %s", error_msg)
-            raise RuntimeError(f"Git clone failed: {error_msg}")
-        logger.debug("Cloned %s (branch: %s) to %s", remote_url, branch, path)
+
+        env, key_path = GitService._prepare_ssh_env(ssh_key)
+        try:
+            proc = await asyncio.create_subprocess_exec(
+                *cmd,
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+                env={**os.environ, **env} if env else None,
+            )
+            stdout, stderr = await proc.communicate()
+            if proc.returncode != 0:
+                error_msg = stderr.decode().strip() if stderr else "unknown error"
+                logger.error("Git clone failed: %s", error_msg)
+                raise RuntimeError(f"Git clone failed: {error_msg}")
+            logger.debug("Cloned %s (branch: %s) to %s", remote_url, branch, path)
+        finally:
+            if key_path and os.path.exists(key_path):
+                os.unlink(key_path)
 
     @staticmethod
-    async def fetch(path: str) -> None:
+    async def fetch(path: str, ssh_key: str | None = None) -> None:
         """Fetch from origin.
 
         Args:
             path: The path to the local git repository.
+            ssh_key: Optional decrypted SSH private key for authentication.
 
         Raises:
             RuntimeError: If fetch fails.
         """
-        proc = await asyncio.create_subprocess_exec(
-            "git",
-            "-C",
-            path,
-            "fetch",
-            "origin",
-            stdout=asyncio.subprocess.PIPE,
-            stderr=asyncio.subprocess.PIPE,
-        )
-        stdout, stderr = await proc.communicate()
-        if proc.returncode != 0:
-            error_msg = stderr.decode().strip() if stderr else "unknown error"
-            logger.error("Git fetch failed: %s", error_msg)
-            raise RuntimeError(f"Git fetch failed: {error_msg}")
-        logger.debug("Fetched origin for %s", path)
+        env, key_path = GitService._prepare_ssh_env(ssh_key)
+        try:
+            proc = await asyncio.create_subprocess_exec(
+                "git",
+                "-C",
+                path,
+                "fetch",
+                "origin",
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+                env={**os.environ, **env} if env else None,
+            )
+            stdout, stderr = await proc.communicate()
+            if proc.returncode != 0:
+                error_msg = stderr.decode().strip() if stderr else "unknown error"
+                logger.error("Git fetch failed: %s", error_msg)
+                raise RuntimeError(f"Git fetch failed: {error_msg}")
+            logger.debug("Fetched origin for %s", path)
+        finally:
+            if key_path and os.path.exists(key_path):
+                os.unlink(key_path)
 
     @staticmethod
-    async def pull(path: str, branch: str) -> None:
+    async def pull(path: str, branch: str, ssh_key: str | None = None) -> None:
         """Pull latest changes from origin.
 
         Args:
             path: The path to the local git repository.
             branch: The branch to pull.
+            ssh_key: Optional decrypted SSH private key for authentication.
 
         Raises:
             RuntimeError: If pull fails.
         """
-        proc = await asyncio.create_subprocess_exec(
-            "git",
-            "-C",
-            path,
-            "pull",
-            "origin",
-            branch,
-            stdout=asyncio.subprocess.PIPE,
-            stderr=asyncio.subprocess.PIPE,
-        )
-        stdout, stderr = await proc.communicate()
-        if proc.returncode != 0:
-            error_msg = stderr.decode().strip() if stderr else "unknown error"
-            logger.error("Git pull failed: %s", error_msg)
-            raise RuntimeError(f"Git pull failed: {error_msg}")
-        logger.debug("Pulled origin/%s for %s", branch, path)
+        env, key_path = GitService._prepare_ssh_env(ssh_key)
+        try:
+            proc = await asyncio.create_subprocess_exec(
+                "git",
+                "-C",
+                path,
+                "pull",
+                "origin",
+                branch,
+                stdout=asyncio.subprocess.PIPE,
+                stderr=asyncio.subprocess.PIPE,
+                env={**os.environ, **env} if env else None,
+            )
+            stdout, stderr = await proc.communicate()
+            if proc.returncode != 0:
+                error_msg = stderr.decode().strip() if stderr else "unknown error"
+                logger.error("Git pull failed: %s", error_msg)
+                raise RuntimeError(f"Git pull failed: {error_msg}")
+            logger.debug("Pulled origin/%s for %s", branch, path)
+        finally:
+            if key_path and os.path.exists(key_path):
+                os.unlink(key_path)
 
     @staticmethod
-    def branch_exists_remotely(path: str, branch: str) -> bool:
+    def branch_exists_remotely(
+        path: str, branch: str, ssh_key: str | None = None
+    ) -> bool:
         """Check if a branch exists on the remote.
 
         Args:
             path: The path to the local git repository.
             branch: The branch name to check.
+            ssh_key: Optional decrypted SSH private key for authentication.
 
         Returns:
             True if the branch exists on origin, False otherwise.
         """
-        result = subprocess.run(
-            ["git", "-C", path, "ls-remote", "--heads", "origin", branch],
-            capture_output=True,
-            text=True,
-        )
-        exists = result.returncode == 0 and result.stdout.strip() != ""
-        logger.debug("Branch %s exists on remote: %s", branch, exists)
-        return exists
+        env, key_path = GitService._prepare_ssh_env(ssh_key)
+        try:
+            result = subprocess.run(
+                ["git", "-C", path, "ls-remote", "--heads", "origin", branch],
+                capture_output=True,
+                text=True,
+                env={**os.environ, **env} if env else None,
+            )
+            exists = result.returncode == 0 and result.stdout.strip() != ""
+            logger.debug("Branch %s exists on remote: %s", branch, exists)
+            return exists
+        finally:
+            if key_path and os.path.exists(key_path):
+                os.unlink(key_path)

apps/api/src/services/manifest_compiler.py

@@ -26,7 +26,7 @@ def resolve_base(manifest: dict) -> dict:
     result = deepcopy(manifest)
 
     base_definition_id = result.pop("base_definition_id", None)
-    base_version = result.pop("base_version", "latest")
+    result.pop("base_version", None)
 
     if base_definition_id:
         # This will be provided by the caller (they have the DB session)
@@ -118,6 +118,11 @@ def compile_dockerfile(manifest: dict) -> str:
 
     # System packages (apt)
     apt_packages = manifest.get("packages", {}).get("apt", [])
+    if manifest.get("user"):
+        # Ensure sudo is available for permission-fixing startup scripts
+        apt_packages = list(apt_packages)
+        if "sudo" not in apt_packages:
+            apt_packages.append("sudo")
     if apt_packages:
         lines.append("RUN apt-get update && apt-get install -y \\")
         for pkg in apt_packages[:-1]:
@@ -167,6 +172,17 @@ def compile_dockerfile(manifest: dict) -> str:
         lines.append(f"ENV HOME={home}")
         lines.append(f"ENV USER={name}")
         lines.append("")
+        # Ensure home directory exists and is writable by the user
+        lines.append(
+            f"RUN mkdir -p {home} && chown {name}:{name} {home} && chmod 755 {home}"
+        )
+        lines.append("")
+
+        # Configure passwordless sudo so startup scripts can fix permissions
+        lines.append(
+            f'RUN echo "{name} ALL=(ALL) NOPASSWD:ALL" > /etc/sudoers.d/{name} && chmod 0440 /etc/sudoers.d/{name}'
+        )
+        lines.append("")
 
     # Build scripts
     build_scripts = manifest.get("scripts", {}).get("build", [])
@@ -181,6 +197,11 @@ def compile_dockerfile(manifest: dict) -> str:
     if build_scripts:
         lines.append("")
 
+    # After build scripts, ensure everything in home is owned by the user
+    if user and build_scripts:
+        lines.append(f"RUN chown -R {name}:{name} {home}")
+        lines.append("")
+
     # Create mount target directories
     mounts = manifest.get("mounts", [])
     if mounts:
@@ -308,7 +329,21 @@ def compile_compose(manifest: dict, variables: dict[str, Any]) -> str:
         service["volumes"] = sort_volumes_by_specificity(volumes)
 
     compose = {"services": {"app": service}}
-    return yaml.dump(compose, default_flow_style=False)
+    result = yaml.dump(compose, default_flow_style=False)
+
+    # Debug: log mount resolution so we can diagnose missing mounts
+    import logging
+    logger = logging.getLogger(__name__)
+    logger.debug(
+        "compile_compose: REPO_PATH=%s SSH_PATH=%s EXTRA_VOLUMES=%s mounts=%s volumes=%s",
+        variables.get("REPO_PATH", "<empty>"),
+        variables.get("SSH_PATH", "<empty>"),
+        variables.get("EXTRA_VOLUMES", []),
+        manifest.get("mounts", []),
+        volumes,
+    )
+
+    return result
 
 
 def resolve_mount_source(mount: dict, variables: dict[str, Any]) -> str:

apps/api/src/services/ssh_keys.py

@@ -2,6 +2,7 @@
 
 import logging
 import os
+import re
 from pathlib import Path
 
 from cryptography.fernet import Fernet
@@ -22,12 +23,28 @@ def _get_fernet() -> Fernet:
     return Fernet(key)
 
 
+def _sanitize_filename(name: str) -> str:
+    """Sanitize a string for use as a filename.
+
+    Replaces non-alphanumeric characters with underscores and strips
+    leading/trailing underscores.
+    """
+    sanitized = re.sub(r"[^a-zA-Z0-9_-]", "_", name)
+    sanitized = sanitized.strip("_")
+    # Ensure it's not empty
+    if not sanitized:
+        sanitized = "key"
+    return sanitized
+
+
 def prepare_ssh_key_files(
     instance_dir: str,
     ssh_key,
     subdir: str = ".ssh",
     uid: int | None = None,
     gid: int | None = None,
+    key_filename: str = "id_ed25519",
+    write_config: bool = True,
 ) -> str:
     """Decrypt and write SSH key files to instance directory for container mounting.
 
@@ -37,6 +54,12 @@ def prepare_ssh_key_files(
         subdir: Subdirectory within instance_dir to write to (default: ".ssh")
         uid: Optional UID to own the files (for bind-mount into non-root container)
         gid: Optional GID to own the files
+        key_filename: Base filename for the key pair (default: "id_ed25519").
+            The private key will be named "{key_filename}" and the public key
+            "{key_filename}.pub".
+        write_config: Whether to write an SSH config file (default: True).
+            Set to False when combining multiple keys into one directory,
+            then call write_ssh_config() separately.
 
     Returns:
         Path to the .ssh directory
@@ -49,51 +72,101 @@ def prepare_ssh_key_files(
     private_key = fernet.decrypt(ssh_key.private_key_encrypted.encode()).decode()
 
     # Write private key with restricted permissions
-    private_key_path = ssh_dir / "id_ed25519"
+    private_key_path = ssh_dir / key_filename
     private_key_path.write_text(private_key)
     os.chmod(private_key_path, 0o600)
 
     # Write public key
-    public_key_path = ssh_dir / "id_ed25519.pub"
+    public_key_path = ssh_dir / f"{key_filename}.pub"
     public_key_path.write_text(ssh_key.public_key)
     os.chmod(public_key_path, 0o644)
 
-    # Write SSH config
-    config_path = ssh_dir / "config"
-    config_content = """Host *
+    # Write SSH config (only if requested)
+    if write_config:
+        config_path = ssh_dir / "config"
+        config_content = f"""Host *
     StrictHostKeyChecking no
     UserKnownHostsFile /dev/null
-    IdentityFile ~/.ssh/id_ed25519
+    IdentityFile ~/.ssh/{key_filename}
     IdentitiesOnly yes
 """
+        config_path.write_text(config_content)
+        os.chmod(config_path, 0o644)
+
+        # Set ownership to target container user if requested
+        if uid is not None or gid is not None:
+            effective_uid = uid if uid is not None else -1
+            effective_gid = gid if gid is not None else -1
+            try:
+                os.chown(ssh_dir, effective_uid, effective_gid)
+                os.chown(private_key_path, effective_uid, effective_gid)
+                os.chown(public_key_path, effective_uid, effective_gid)
+                os.chown(config_path, effective_uid, effective_gid)
+                logger.debug(
+                    "Set SSH key ownership to uid=%s gid=%s for %s",
+                    effective_uid,
+                    effective_gid,
+                    ssh_dir,
+                )
+            except PermissionError as exc:
+                logger.warning(
+                    "Cannot chown SSH keys to uid=%s gid=%s (running as uid=%s): %s",
+                    effective_uid,
+                    effective_gid,
+                    os.getuid(),
+                    exc,
+                )
+    else:
+        # Still chown the key files even if we didn't write config
+        if uid is not None or gid is not None:
+            effective_uid = uid if uid is not None else -1
+            effective_gid = gid if gid is not None else -1
+            try:
+                os.chown(private_key_path, effective_uid, effective_gid)
+                os.chown(public_key_path, effective_uid, effective_gid)
+            except PermissionError:
+                pass
+
+    return str(ssh_dir)
+
+
+def write_ssh_config(
+    ssh_dir: str,
+    key_filenames: list[str],
+    uid: int | None = None,
+    gid: int | None = None,
+) -> None:
+    """Write an SSH config file that includes multiple IdentityFile entries.
+
+    Args:
+        ssh_dir: Path to the .ssh directory
+        key_filenames: List of key filenames (without .pub extension)
+        uid: Optional UID to own the config file
+        gid: Optional GID to own the config file
+    """
+    ssh_dir_path = Path(ssh_dir)
+    ssh_dir_path.mkdir(parents=True, exist_ok=True)
+
+    config_path = ssh_dir_path / "config"
+    lines = ["Host *"]
+    lines.append("    StrictHostKeyChecking no")
+    lines.append("    UserKnownHostsFile /dev/null")
+    lines.append("    IdentitiesOnly yes")
+    for filename in key_filenames:
+        lines.append(f"    IdentityFile ~/.ssh/{filename}")
+    lines.append("")
+
+    config_content = "\n".join(lines)
     config_path.write_text(config_content)
     os.chmod(config_path, 0o644)
 
-    # Set ownership to target container user if requested
     if uid is not None or gid is not None:
         effective_uid = uid if uid is not None else -1
         effective_gid = gid if gid is not None else -1
         try:
-            os.chown(ssh_dir, effective_uid, effective_gid)
-            os.chown(private_key_path, effective_uid, effective_gid)
-            os.chown(public_key_path, effective_uid, effective_gid)
             os.chown(config_path, effective_uid, effective_gid)
-            logger.debug(
-                "Set SSH key ownership to uid=%s gid=%s for %s",
-                effective_uid,
-                effective_gid,
-                ssh_dir,
-            )
-        except PermissionError as exc:
-            logger.warning(
-                "Cannot chown SSH keys to uid=%s gid=%s (running as uid=%s): %s",
-                effective_uid,
-                effective_gid,
-                os.getuid(),
-                exc,
-            )
-
-    return str(ssh_dir)
+        except PermissionError:
+            pass
 
 
 def cleanup_ssh_key_files(instance_dir: str) -> None:

apps/api/src/services/terminal_manager.py

@@ -6,6 +6,7 @@ import uuid
 from datetime import datetime, timezone
 
 from fastapi import WebSocket
+from sqlalchemy.dialects.postgresql import insert as pg_insert
 
 from src.database import SessionLocal
 from src.models.terminal_session import TerminalSessionModel
@@ -83,18 +84,26 @@ class TerminalManager:
         instance_id: uuid.UUID,
         name: str,
     ) -> None:
-        """Insert a TerminalSessionModel row into the database."""
+        """Insert a TerminalSessionModel row into the database.
+
+        Uses ON CONFLICT DO NOTHING to handle races when a session is
+        restored from DB and then re-inserted.
+        """
         try:
             async with SessionLocal() as db_session:
-                db_row = TerminalSessionModel(
-                    id=uuid.UUID(session_id),
-                    instance_id=instance_id,
-                    name=name,
-                    status="active",
-                    created_at=datetime.now(timezone.utc),
-                    last_activity_at=datetime.now(timezone.utc),
+                stmt = (
+                    pg_insert(TerminalSessionModel)
+                    .values(
+                        id=uuid.UUID(session_id),
+                        instance_id=instance_id,
+                        name=name,
+                        status="active",
+                        created_at=datetime.now(timezone.utc),
+                        last_activity_at=datetime.now(timezone.utc),
+                    )
+                    .on_conflict_do_nothing(index_elements=["id"])
                 )
-                db_session.add(db_row)
+                await db_session.execute(stmt)
                 await db_session.commit()
                 logger.debug(
                     "Inserted terminal session row %s for instance %s",

apps/api/src/services/terminal_session.py

@@ -1,10 +1,13 @@
-"""Terminal session management for tool instances."""
+"""High-performance terminal session with asyncio-native I/O.
+
+Replaces blocking select.select() with event-driven asyncio.add_reader()
+for sub-frame latency. Includes output batching and flow control.
+"""
 
 import asyncio
 import logging
 import os
 import pty
-import select
 import signal
 import struct
 import fcntl
@@ -17,18 +20,31 @@ logger = logging.getLogger(__name__)
 
 
 class TerminalSession:
-    """Manages a single terminal session connected to a docker container.
+    """Manages a single terminal session with event-driven PTY I/O.
 
-    Supports persistent sessions that survive WebSocket disconnections.
-    Multiple WebSocket connections can attach/detach from the same session.
+    Uses asyncio.add_reader() instead of polling for near-zero read latency.
+    Output is batched (2ms window) and sent as binary WebSocket frames.
+    Flow control prevents memory bloat on fast output.
     """
 
-    # Circular buffer size (10KB)
+    # Circular buffer for replay (10KB)
     BUFFER_SIZE = 10 * 1024
 
     # Idle timeout in seconds (30 minutes)
     IDLE_TIMEOUT = 30 * 60
 
+    # Output batching window in seconds
+    BATCH_WINDOW_S = 0.002  # 2ms
+
+    # Flow control: pause PTY reads when unacknowledged bytes exceed this
+    FLOW_CONTROL_PAUSE = 64 * 1024
+
+    # Flow control: resume PTY reads when unacknowledged bytes drop below this
+    FLOW_CONTROL_RESUME = 32 * 1024
+
+    # Max WebSocket frame size
+    MAX_FRAME_SIZE = 64 * 1024
+
     # Session number counter per instance_id for auto-naming
     _instance_counters: dict[str, int] = {}
 
@@ -47,7 +63,6 @@ class TerminalSession:
         self.process: asyncio.subprocess.Process | None = None
         self._closed = False
         self._master_fd: int | None = None
-        self._slave_fd: int | None = None
 
         # Circular buffer for output replay
         self._output_buffer: deque[bytes] = deque(maxlen=self.BUFFER_SIZE)
@@ -67,6 +82,20 @@ class TerminalSession:
         self.name = name or self._generate_name(str(instance_id))
         self.status: str = "active"
 
+        # Output batching
+        self._batch_buffer = bytearray()
+        self._batch_timer: asyncio.TimerHandle | None = None
+        self._batch_lock = asyncio.Lock()
+
+        # Flow control
+        self._unacknowledged_bytes = 0
+        self._paused = False
+        self._read_handler_set = False
+        self._flow_control_lock = asyncio.Lock()
+
+        # Ack timeout fallback
+        self._ack_timeout_handle: asyncio.TimerHandle | None = None
+
     @classmethod
     def _generate_name(cls, instance_id: str) -> str:
         """Generate an auto-incremented session name for the instance."""
@@ -77,91 +106,225 @@ class TerminalSession:
     async def start(self, startup_command: str | None = None) -> None:
         """Start the docker exec process with a shell using a PTY."""
         # Create a pseudo-terminal on the host
-        self._master_fd, self._slave_fd = pty.openpty()
+        self._master_fd, slave_fd = pty.openpty()
 
         # Set the terminal size initially
         self._set_terminal_size(self._cols, self._rows)
         logger.debug(
-            f"Starting terminal session {self.session_id} for container {self.container_id} with initial size {self._cols}x{self._rows}"
+            "Starting terminal session %s for container %s with initial size %sx%s",
+            self.session_id,
+            self.container_id,
+            self._cols,
+            self._rows,
         )
 
         # Build the shell command
-        if startup_command:
-            shell_cmd = f'bash -c "{startup_command}" || true; exec bash -il'
+        cmd = startup_command or self.startup_command
+        if cmd:
+            shell_cmd = f'bash -c "{cmd}" || true; exec bash -il'
             logger.debug(
-                f"Using startup command for session {self.session_id}: {startup_command}"
+                "Using startup command for session %s: %s",
+                self.session_id,
+                cmd,
             )
         else:
             shell_cmd = "bash -il"
 
         # Start docker exec with the slave fd as stdin/stdout/stderr
-        # Using -it because the slave fd IS a TTY
         self.process = await asyncio.create_subprocess_exec(
             "docker",
             "exec",
             "-it",
             "-e",
-            "TERM=xterm",
+            "TERM=xterm-256color",
             self.container_id,
             "bash",
             "-c",
             shell_cmd,
-            stdin=self._slave_fd,
-            stdout=self._slave_fd,
-            stderr=self._slave_fd,
+            stdin=slave_fd,
+            stdout=slave_fd,
+            stderr=slave_fd,
         )
 
         # Close slave fd in parent process
-        os.close(self._slave_fd)
-        self._slave_fd = None
+        os.close(slave_fd)
 
         self.last_activity = time.time()
 
-    def _set_terminal_size(self, cols: int, rows: int) -> None:
-        """Set the terminal size using TIOCSWINSZ."""
-        if self._master_fd is None:
-            logger.warning("Cannot resize: master_fd is None (session not started)")
-            return
-        # TIOCSWINSZ = 0x5414 on Linux
-        TIOCSWINSZ = 0x5414
-        size = struct.pack("HHHH", rows, cols, 0, 0)
-        try:
-            fcntl.ioctl(self._master_fd, TIOCSWINSZ, size)
-            logger.debug(f"Resized PTY to {cols}x{rows} (fd={self._master_fd})")
-        except (OSError, IOError) as e:
-            logger.error(f"Failed to resize PTY: {e}")
+        # Start event-driven reading
+        self._start_reading()
 
-    async def read_output(self) -> bytes:
-        """Read output from the PTY master and store in buffer."""
-        if self._master_fd is None or self._closed:
-            return b""
+    def _start_reading(self) -> None:
+        """Register PTY master fd with asyncio event loop for event-driven reads."""
+        if self._read_handler_set or self._master_fd is None or self._closed:
+            return
         try:
-            # Use select to check if data is available
-            readable, _, _ = select.select([self._master_fd], [], [], 0.1)
-            if readable:
-                data = os.read(self._master_fd, 4096)
-                if data:
-                    self._add_to_buffer(data)
-                    self.last_activity = time.time()
-                return data
-            return b""
-        except (OSError, IOError, ValueError):
-            return b""
+            loop = asyncio.get_event_loop()
+            loop.add_reader(self._master_fd, self._on_fd_readable)
+            self._read_handler_set = True
+            logger.debug("Started event-driven reading for session %s", self.session_id)
+        except Exception as exc:
+            logger.error(
+                "Failed to start reading for session %s: %s", self.session_id, exc
+            )
+
+    def _stop_reading(self) -> None:
+        """Unregister PTY master fd from asyncio event loop."""
+        if not self._read_handler_set or self._master_fd is None:
+            return
+        try:
+            loop = asyncio.get_event_loop()
+            loop.remove_reader(self._master_fd)
+            self._read_handler_set = False
+        except Exception:
+            pass
+
+    def _on_fd_readable(self) -> None:
+        """Callback when PTY master fd has data available (called by event loop)."""
+        if self._master_fd is None or self._closed:
+            return
+
+        try:
+            data = os.read(self._master_fd, 4096)
+        except (OSError, IOError) as exc:
+            logger.debug("PTY read error for session %s: %s", self.session_id, exc)
+            self._handle_eof()
+            return
+
+        if not data:
+            # EOF: docker exec process exited
+            logger.debug("PTY EOF for session %s", self.session_id)
+            self._handle_eof()
+            return
+
+        self._add_to_buffer(data)
+        self.last_activity = time.time()
+
+        # Queue for batching + flow control
+        self._queue_output(data)
 
     def _add_to_buffer(self, data: bytes) -> None:
         """Add data to circular buffer, maintaining size limit."""
         self._output_buffer.append(data)
         self._buffer_size += len(data)
-
-        # Trim if exceeds max size
         while self._buffer_size > self.BUFFER_SIZE and self._output_buffer:
             removed = self._output_buffer.popleft()
             self._buffer_size -= len(removed)
 
+    def _queue_output(self, data: bytes) -> None:
+        """Add output to batch buffer and schedule flush."""
+        self._batch_buffer.extend(data)
+        self._unacknowledged_bytes += len(data)
+
+        # Check flow control
+        if self._unacknowledged_bytes > self.FLOW_CONTROL_PAUSE and not self._paused:
+            self._pause_output()
+
+        # Schedule batch flush if not already scheduled
+        if self._batch_timer is None:
+            loop = asyncio.get_event_loop()
+            self._batch_timer = loop.call_later(
+                self.BATCH_WINDOW_S,
+                self._flush_batch_sync,
+            )
+
+    def _flush_batch_sync(self) -> None:
+        """Synchronous entry point for batch flush (called from event loop)."""
+        self._batch_timer = None
+        if not self._batch_buffer or not self._websockets:
+            self._batch_buffer.clear()
+            return
+
+        payload = bytes(self._batch_buffer)
+        self._batch_buffer.clear()
+
+        # Send to all websockets (asyncio.create_task for async send)
+        dead_sockets = set()
+        for ws in list(self._websockets):
+            try:
+                asyncio.create_task(self._send_bytes(ws, payload))
+            except Exception:
+                dead_sockets.add(ws)
+
+        if dead_sockets:
+            self._websockets -= dead_sockets
+
+    async def _send_bytes(self, ws: Any, payload: bytes) -> None:
+        """Send bytes to a single websocket, catching errors."""
+        try:
+            await ws.send_bytes(payload)
+        except Exception:
+            self._websockets.discard(ws)
+
+    def acknowledge_data(self, char_count: int) -> None:
+        """Client acknowledges processing char_count bytes.
+
+        Called from the WebSocket handler when the client sends an 'ack' message.
+        """
+        self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
+
+        if self._paused and self._unacknowledged_bytes < self.FLOW_CONTROL_RESUME:
+            self._resume_output()
+
+        # Reset ack timeout
+        if self._ack_timeout_handle:
+            self._ack_timeout_handle.cancel()
+        loop = asyncio.get_event_loop()
+        self._ack_timeout_handle = loop.call_later(5.0, self._ack_timeout_fallback)
+
+    def _ack_timeout_fallback(self) -> None:
+        """If no ack received for 5s, assume client is dead and resume."""
+        logger.warning(
+            "Flow control ack timeout for session %s, resuming output",
+            self.session_id,
+        )
+        self._unacknowledged_bytes = 0
+        if self._paused:
+            self._resume_output()
+
+    def _pause_output(self) -> None:
+        """Pause reading from PTY due to flow control."""
+        self._paused = True
+        self._stop_reading()
+        logger.debug(
+            "Paused output for session %s (%d unacked)",
+            self.session_id,
+            self._unacknowledged_bytes,
+        )
+
+    def _resume_output(self) -> None:
+        """Resume reading from PTY."""
+        self._paused = False
+        self._start_reading()
+        logger.debug("Resumed output for session %s", self.session_id)
+
     def get_buffer(self) -> bytes:
         """Get buffered output for replay."""
         return b"".join(self._output_buffer)
 
+    def _handle_eof(self) -> None:
+        """Handle PTY EOF: process died, close websockets to force reconnect."""
+        self._stop_reading()
+        # Mark process as done so is_alive() returns False
+        if self.process is not None and self.process.returncode is None:
+            # Force returncode to a non-None value since the process is dead
+            # but asyncio.subprocess may not have set it yet
+            try:
+                self.process._transport.close()  # type: ignore[attr-defined]
+            except Exception:
+                pass
+        # Close all websockets to force frontend reconnection
+        dead_sockets = set(self._websockets)
+        self._websockets.clear()
+        for ws in dead_sockets:
+            try:
+                asyncio.create_task(
+                    ws.close(code=4001, reason="Session process exited")
+                )
+            except Exception:
+                pass
+        logger.info("Session %s EOF handled, websockets closed", self.session_id)
+
     async def write_input(self, data: bytes) -> None:
         """Write input to the PTY master."""
         if self._master_fd is None or self._closed:
@@ -169,8 +332,22 @@ class TerminalSession:
         try:
             os.write(self._master_fd, data)
             self.last_activity = time.time()
-        except (OSError, IOError):
-            pass
+        except (OSError, IOError) as exc:
+            logger.debug("PTY write error for session %s: %s", self.session_id, exc)
+            self._handle_eof()
+
+    def _set_terminal_size(self, cols: int, rows: int) -> None:
+        """Set the terminal size using TIOCSWINSZ."""
+        if self._master_fd is None:
+            logger.warning("Cannot resize: master_fd is None (session not started)")
+            return
+        TIOCSWINSZ = 0x5414
+        size = struct.pack("HHHH", rows, cols, 0, 0)
+        try:
+            fcntl.ioctl(self._master_fd, TIOCSWINSZ, size)
+            logger.debug("Resized PTY to %sx%s (fd=%s)", cols, rows, self._master_fd)
+        except (OSError, IOError) as e:
+            logger.error("Failed to resize PTY: %s", e)
 
     async def resize(self, cols: int, rows: int) -> None:
         """Resize the terminal."""
@@ -178,32 +355,24 @@ class TerminalSession:
             logger.warning("Cannot resize: session is closed")
             return
 
-        # Only resize if dimensions actually changed
         if cols == self._cols and rows == self._rows:
             return
 
         self._cols = cols
         self._rows = rows
-        logger.debug(f"resize() called for session {self.session_id}: {cols}x{rows}")
+        logger.debug(
+            "resize() called for session %s: %sx%s", self.session_id, cols, rows
+        )
         self._set_terminal_size(cols, rows)
 
-        # Docker exec -it creates its own PTY inside the container,
-        # so host PTY resize doesn't propagate to the container shell.
-        # Send SIGWINCH to the docker exec process on the host.
-        # Docker exec forwards signals to the container process, which should
-        # cause the container's shell to re-read its terminal size.
+        # Send SIGWINCH to docker exec process
         if self.process and self.process.pid:
             try:
                 os.kill(self.process.pid, signal.SIGWINCH)
-                logger.debug(
-                    f"Sent SIGWINCH to docker exec process {self.process.pid} for session {self.session_id}"
-                )
             except ProcessLookupError:
-                logger.warning(
-                    f"docker exec process {self.process.pid} not found for session {self.session_id}"
-                )
+                logger.warning("docker exec process %s not found", self.process.pid)
             except Exception as e:
-                logger.warning(f"Failed to send SIGWINCH: {e}")
+                logger.warning("Failed to send SIGWINCH: %s", e)
 
     async def reset(self) -> None:
         """Reset the session by killing the process and clearing state."""
@@ -213,9 +382,13 @@ class TerminalSession:
         self._output_buffer.clear()
         self._buffer_size = 0
         self._websockets.clear()
+        self._batch_buffer.clear()
+        self._batch_timer = None
+        self._unacknowledged_bytes = 0
+        self._paused = False
+        self._read_handler_set = False
         self.process = None
         self._master_fd = None
-        self._slave_fd = None
         self.status = "active"
 
     async def close(self) -> None:
@@ -225,11 +398,21 @@ class TerminalSession:
         self._closed = True
         self.status = "closed"
 
+        self._stop_reading()
+
+        if self._batch_timer:
+            self._batch_timer.cancel()
+            self._batch_timer = None
+
+        if self._ack_timeout_handle:
+            self._ack_timeout_handle.cancel()
+            self._ack_timeout_handle = None
+
         if self._master_fd is not None:
             try:
                 os.close(self._master_fd)
             except OSError:
-                pass  # noqa: S110
+                pass
             self._master_fd = None
 
         if self.process is not None:
@@ -265,14 +448,20 @@ class TerminalSession:
         return len(self._websockets) > 0
 
     async def send_to_all(self, data: bytes) -> None:
-        """Send data to all attached WebSockets."""
+        """Send data to all attached WebSockets (used for control messages)."""
         dead_sockets = set()
         for ws in self._websockets:
             try:
                 await ws.send_bytes(data)
             except Exception:
                 dead_sockets.add(ws)
-
-        # Clean up dead sockets
         for ws in dead_sockets:
             self._websockets.discard(ws)
+
+    async def read_output(self) -> bytes:
+        """Legacy method: read output synchronously.
+
+        With event-driven I/O, output is automatically sent to websockets.
+        This method returns any buffered data for callers that poll.
+        """
+        return b""

apps/api/src/services/workspace_manager.py

@@ -2,9 +2,11 @@
 
 from __future__ import annotations
 
+import contextlib
 import logging
 import os
 import shutil
+import stat
 import uuid
 from dataclasses import dataclass
 from datetime import datetime
@@ -14,6 +16,7 @@ from sqlalchemy import select
 
 from src.models.workspace import Workspace
 from src.services.git_service import GitService
+from src.services.ssh_keys import _get_fernet
 
 if TYPE_CHECKING:
     from sqlalchemy.ext.asyncio import AsyncSession
@@ -34,7 +37,7 @@ class SyncResult:
 class WorkspaceHasInstancesError(Exception):
     """Raised when attempting to delete a workspace with running instances."""
 
-    def __init__(self, instances: list[ToolInstance]) -> None:
+    def __init__(self, instances: list[dict]) -> None:
         self.instances = instances
         super().__init__(f"Workspace has {len(instances)} running tool instance(s)")
 
@@ -54,6 +57,7 @@ class WorkspaceManager:
         user_id: uuid.UUID,
         name: str,
         branch: str = "main",
+        session: AsyncSession | None = None,
     ) -> Workspace:
         """Clone repo to workspace path and create DB record.
 
@@ -62,6 +66,7 @@ class WorkspaceManager:
             user_id: The owner user ID.
             name: The workspace name (unique per repo).
             branch: The branch to clone (default: "main").
+            session: Database session for loading SSH keys.
 
         Returns:
             The created Workspace record.
@@ -70,7 +75,11 @@ class WorkspaceManager:
             RuntimeError: If git clone fails.
         """
         path = self._workspace_path(repo.id, name)
-        os.makedirs(os.path.dirname(path), exist_ok=True)
+        parent = os.path.dirname(path)
+        os.makedirs(parent, exist_ok=True)
+        # Ensure container users (various UIDs) can write to workspace dirs
+        with contextlib.suppress(OSError):
+            os.chmod(parent, 0o777)
 
         logger.info(
             "Creating workspace: name=%s, repo=%s, branch=%s", name, repo.id, branch
@@ -79,7 +88,28 @@ class WorkspaceManager:
         if not repo.remote_url:
             raise ValueError("Repository has no remote URL")
 
-        await GitService.clone(repo.remote_url, branch, path)
+        # Remove stale directory from previous failed/aborted clone
+        if os.path.exists(path):
+            logger.warning("Removing stale workspace directory: %s", path)
+            shutil.rmtree(path, ignore_errors=True)
+
+        # Load SSH key if repo has one
+        ssh_key = None
+        if getattr(repo, "ssh_key_id", None) and session is not None:
+            from src.models.ssh_key import SSHKey
+
+            result = await session.execute(
+                select(SSHKey).where(SSHKey.id == repo.ssh_key_id)
+            )
+            ssh_key_obj = result.scalar_one_or_none()
+            if ssh_key_obj:
+                fernet = _get_fernet()
+                ssh_key = fernet.decrypt(
+                    ssh_key_obj.private_key_encrypted.encode()
+                ).decode()
+
+        await GitService.clone(repo.remote_url, branch, path, ssh_key=ssh_key)
+        self._make_world_writable(path)
 
         workspace = Workspace(
             name=name,
@@ -114,7 +144,9 @@ class WorkspaceManager:
 
         instances = await self._get_instances(workspace, session)
         if instances and not force:
-            raise WorkspaceHasInstancesError(instances)
+            raise WorkspaceHasInstancesError(
+                [{"id": str(i.id), "name": i.name} for i in instances]
+            )
 
         # Stop and delete all instances
         for instance in instances:
@@ -129,11 +161,14 @@ class WorkspaceManager:
         await session.delete(workspace)
         logger.info("Deleted workspace record: %s", workspace.id)
 
-    async def sync(self, workspace: Workspace) -> SyncResult:
+    async def sync(
+        self, workspace: Workspace, session: AsyncSession | None = None
+    ) -> SyncResult:
         """Sync a workspace with its remote.
 
         Args:
             workspace: The workspace to sync.
+            session: Database session for loading SSH keys.
 
         Returns:
             SyncResult indicating whether the branch was deleted.
@@ -143,16 +178,63 @@ class WorkspaceManager:
         """
         logger.info("Syncing workspace: %s", workspace.id)
 
-        await GitService.fetch(workspace.path)
+        # Load SSH key if repo has one
+        ssh_key = None
+        if session is not None:
+            from src.models.git_repository import GitRepository
+            from src.models.ssh_key import SSHKey
 
-        if not GitService.branch_exists_remotely(workspace.path, workspace.branch):
+            repo = await session.get(GitRepository, workspace.repo_id)
+            if repo and getattr(repo, "ssh_key_id", None):
+                result = await session.execute(
+                    select(SSHKey).where(SSHKey.id == repo.ssh_key_id)
+                )
+                ssh_key_obj = result.scalar_one_or_none()
+                if ssh_key_obj:
+                    fernet = _get_fernet()
+                    ssh_key = fernet.decrypt(
+                        ssh_key_obj.private_key_encrypted.encode()
+                    ).decode()
+
+        await GitService.fetch(workspace.path, ssh_key=ssh_key)
+
+        if not GitService.branch_exists_remotely(
+            workspace.path, workspace.branch, ssh_key=ssh_key
+        ):
             return SyncResult(branch_deleted=True)
 
-        await GitService.pull(workspace.path, workspace.branch)
+        await GitService.pull(workspace.path, workspace.branch, ssh_key=ssh_key)
+        self._make_world_writable(workspace.path)
+
         workspace.last_sync_at = datetime.now()
         logger.info("Workspace synced: %s", workspace.id)
         return SyncResult(branch_deleted=False)
 
+    def _make_world_writable(self, path: str) -> None:
+        """Recursively make path readable/writable/traversable by any UID.
+
+        Directories get 777 (traversable). Files get rw for all while
+        preserving any existing execute bits.
+        """
+        with contextlib.suppress(OSError):
+            os.chmod(path, 0o777)
+        for root, dirs, files in os.walk(path):
+            for d in dirs:
+                dpath = os.path.join(root, d)
+                with contextlib.suppress(OSError):
+                    os.chmod(dpath, 0o777)
+            for f in files:
+                fpath = os.path.join(root, f)
+                with contextlib.suppress(OSError):
+                    mode = os.stat(fpath).st_mode
+                    # Preserve execute bits, ensure read+write for all
+                    new_mode = (mode & stat.S_IXUSR) | 0o666
+                    if mode & stat.S_IXGRP:
+                        new_mode |= stat.S_IXGRP
+                    if mode & stat.S_IXOTH:
+                        new_mode |= stat.S_IXOTH
+                    os.chmod(fpath, new_mode)
+
     async def _get_instances(
         self,
         workspace: Workspace,

apps/api/tests/unit/test_tool_instances_legacy.py

@@ -145,10 +145,12 @@ class TestCreateInstanceDockerfileLegacy:
         data = MagicMock()
         data.tool_type_id = str(fake_tool_type_id)
         data.display_name = None
+        data.workspace_id = None
         data.clone_mode = "mount"
         data.branch = None
         data.new_branch = None
         data.config_profile_id = None
+        data.ssh_key_ids = []
 
         result = await create_instance(
             project_id=fake_project_id,
@@ -225,10 +227,12 @@ class TestCreateInstanceDockerfileLegacy:
         data = MagicMock()
         data.tool_type_id = str(fake_tool_type_id)
         data.display_name = None
+        data.workspace_id = None
         data.clone_mode = "mount"
         data.branch = None
         data.new_branch = None
         data.config_profile_id = None
+        data.ssh_key_ids = []
 
         with pytest.raises(HTTPException) as exc_info:
             await create_instance(
@@ -305,10 +309,12 @@ class TestCreateInstanceComposeLegacy:
         data = MagicMock()
         data.tool_type_id = str(fake_tool_type_id)
         data.display_name = None
+        data.workspace_id = None
         data.clone_mode = "mount"
         data.branch = None
         data.new_branch = None
         data.config_profile_id = None
+        data.ssh_key_ids = []
 
         result = await create_instance(
             project_id=fake_project_id,
@@ -389,10 +395,12 @@ class TestCreateInstanceManifestNotCalledForLegacy:
         data = MagicMock()
         data.tool_type_id = str(fake_tool_type_id)
         data.display_name = None
+        data.workspace_id = None
         data.clone_mode = "mount"
         data.branch = None
         data.new_branch = None
         data.config_profile_id = None
+        data.ssh_key_ids = []
 
         await create_instance(
             project_id=fake_project_id,
@@ -411,8 +419,10 @@ class TestStartInstanceLegacyFallback:
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._prepare_manifest_instance")
     @patch("src.api.tool_instances._get_user")
@@ -423,8 +433,10 @@ class TestStartInstanceLegacyFallback:
         mock_get_user,
         mock_prepare_manifest,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
@@ -440,7 +452,6 @@ class TestStartInstanceLegacyFallback:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,
@@ -509,8 +520,10 @@ class TestStartInstanceLegacyFallback:
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._prepare_manifest_instance")
     @patch("src.api.tool_instances._get_user")
@@ -521,8 +534,10 @@ class TestStartInstanceLegacyFallback:
         mock_get_user,
         mock_prepare_manifest,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
@@ -538,7 +553,6 @@ class TestStartInstanceLegacyFallback:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,
@@ -606,8 +620,10 @@ class TestStartInstanceLegacyFallback:
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._prepare_manifest_instance")
     @patch("src.api.tool_instances._get_user")
@@ -618,8 +634,10 @@ class TestStartInstanceLegacyFallback:
         mock_get_user,
         mock_prepare_manifest,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
@@ -635,7 +653,6 @@ class TestStartInstanceLegacyFallback:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,
@@ -705,12 +722,15 @@ class TestStartInstanceSshPermissions:
     """SSH key mounts trigger permission fixes after container starts."""
 
     @patch("src.api.tool_instances.write_compose_file")
+    @patch("src.api.tool_instances.prepare_ssh_key_files")
     @patch("src.api.tool_instances.apply_ssh_permissions")
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._get_user")
     @patch("src.api.tool_instances._get_owned_project")
@@ -719,12 +739,15 @@ class TestStartInstanceSshPermissions:
         mock_get_project,
         mock_get_user,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
         mock_apply_ssh,
+        mock_prepare_ssh,
         mock_write_compose,
         mock_session,
         fake_user_id,
@@ -743,7 +766,6 @@ class TestStartInstanceSshPermissions:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,
@@ -815,33 +837,37 @@ class TestStartInstanceSshPermissions:
         mock_session.get.side_effect = _get
 
         with patch("os.path.exists", return_value=True):
-            with patch(
-                "src.api.tool_instances._prepare_manifest_instance"
-            ) as mock_prepare:
-                mock_prepare.return_value = (
-                    "headquarter/test:latest",
-                    "services:\n  app:\n    image: test",
-                    {"name": "test-manifest", "user": {"name": "user"}},
-                    "/home/user",
-                )
-                result = await start_instance(
-                    project_id=fake_project_id,
-                    repo_id=fake_repo_id,
-                    instance_id=fake_instance_id,
-                    data=None,
-                    user_id=fake_user_id,
-                    session=mock_session,
-                )
+            with patch("os.makedirs"):
+                with patch(
+                    "src.api.tool_instances._prepare_manifest_instance"
+                ) as mock_prepare:
+                    mock_prepare.return_value = (
+                        "headquarter/test:latest",
+                        "services:\n  app:\n    image: test",
+                        {"name": "test-manifest", "user": {"name": "user"}},
+                        "/home/user",
+                    )
+                    result = await start_instance(
+                        project_id=fake_project_id,
+                        repo_id=fake_repo_id,
+                        instance_id=fake_instance_id,
+                        data=None,
+                        user_id=fake_user_id,
+                        session=mock_session,
+                    )
 
         assert result["status"] == "running"
         mock_apply_ssh.assert_called_once_with("abc123", "/home/user/.ssh", "user")
 
+    @patch("src.api.tool_instances.prepare_ssh_key_files")
     @patch("src.api.tool_instances.apply_ssh_permissions")
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._get_user")
     @patch("src.api.tool_instances._get_owned_project")
@@ -850,12 +876,15 @@ class TestStartInstanceSshPermissions:
         mock_get_project,
         mock_get_user,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
         mock_apply_ssh,
+        mock_prepare_ssh,
         mock_session,
         fake_user_id,
         fake_project_id,
@@ -870,7 +899,6 @@ class TestStartInstanceSshPermissions:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,
@@ -933,14 +961,16 @@ class TestStartInstanceSshPermissions:
         mock_session.get.side_effect = _get
 
         with patch("os.path.exists", return_value=True):
-            result = await start_instance(
-                project_id=fake_project_id,
-                repo_id=fake_repo_id,
-                instance_id=fake_instance_id,
-                data=None,
-                user_id=fake_user_id,
-                session=mock_session,
-            )
+            with patch("os.makedirs"):
+                with patch("src.api.tool_instances._modify_compose_file"):
+                    result = await start_instance(
+                        project_id=fake_project_id,
+                        repo_id=fake_repo_id,
+                        instance_id=fake_instance_id,
+                        data=None,
+                        user_id=fake_user_id,
+                        session=mock_session,
+                    )
 
         assert result["status"] == "running"
         mock_apply_ssh.assert_called_once_with("abc123", "/root/.ssh", "root")
@@ -952,8 +982,10 @@ class TestStartInstanceManifestBranch:
     @patch("src.api.tool_instances.wait_for_container_running")
     @patch("src.api.tool_instances.execute_compose_command")
     @patch("src.api.tool_instances.get_container_id")
-    @patch("src.api.tool_instances.get_container_name")
     @patch("src.api.tool_instances.connect_container_to_network")
+    @patch("src.api.tool_instances._ensure_backend_network_in_compose")
+    @patch("src.api.tool_instances._ensure_container_name_in_compose")
+    @patch("src.api.tool_instances._ensure_web_bind_address")
     @patch("src.api.tool_instances._sanitize_compose_file")
     @patch("src.api.tool_instances._prepare_manifest_instance")
     @patch("src.api.tool_instances.write_compose_file")
@@ -966,8 +998,10 @@ class TestStartInstanceManifestBranch:
         mock_write_compose,
         mock_prepare_manifest,
         mock_sanitize,
+        mock_ensure_web_bind,
+        mock_ensure_container_name,
+        mock_backend_network,
         mock_connect_network,
-        mock_get_container_name,
         mock_get_container_id,
         mock_execute_compose,
         mock_wait_container,
@@ -987,7 +1021,6 @@ class TestStartInstanceManifestBranch:
         mock_get_project.return_value = AsyncMock()
         mock_execute_compose.return_value = (0, "started", "")
         mock_get_container_id.return_value = "abc123"
-        mock_get_container_name.return_value = "test-container"
         mock_connect_network.return_value = True
         mock_wait_container.return_value = {
             "success": True,

apps/web/nginx.conf

@@ -15,6 +15,12 @@ server {
         try_files $uri $uri/ /index.html;
     }
 
+    # Never cache index.html so browsers always fetch new hashed JS/CSS
+    location = /index.html {
+        add_header Cache-Control "no-cache, no-store, must-revalidate";
+        add_header Pragma "no-cache";
+    }
+
     # Cache static assets
     location ~* \.(js|css|png|jpg|jpeg|gif|ico|svg|woff|woff2)$ {
         expires 1y;

apps/web/package-lock.json

@@ -16,11 +16,11 @@
         "react-dom": "^18.2.0",
         "react-router-dom": "^6.20.0",
         "react-simple-code-editor": "^0.14.1",
-        "sonner": "^1.7.4",
         "tailwindcss": "^3.3.0",
         "xterm": "^5.3.0",
         "xterm-addon-fit": "^0.8.0",
-        "xterm-addon-web-links": "^0.9.0"
+        "xterm-addon-web-links": "^0.9.0",
+        "xterm-addon-webgl": "^0.16.0"
       },
       "devDependencies": {
         "@testing-library/jest-dom": "^6.9.1",
@@ -5469,16 +5469,6 @@
         "node": ">=8"
       }
     },
-    "node_modules/sonner": {
-      "version": "1.7.4",
-      "resolved": "https://registry.npmjs.org/sonner/-/sonner-1.7.4.tgz",
-      "integrity": "sha512-DIS8z4PfJRbIyfVFDVnK9rO3eYDtse4Omcm6bt0oEr5/jtLgysmjuBl1frJ9E/EQZrFmKx2A8m/s5s9CRXIzhw==",
-      "license": "MIT",
-      "peerDependencies": {
-        "react": "^18.0.0 || ^19.0.0 || ^19.0.0-rc",
-        "react-dom": "^18.0.0 || ^19.0.0 || ^19.0.0-rc"
-      }
-    },
     "node_modules/source-map-js": {
       "version": "1.2.1",
       "resolved": "https://registry.npmjs.org/source-map-js/-/source-map-js-1.2.1.tgz",
@@ -6324,6 +6314,16 @@
         "xterm": "^5.0.0"
       }
     },
+    "node_modules/xterm-addon-webgl": {
+      "version": "0.16.0",
+      "resolved": "https://registry.npmjs.org/xterm-addon-webgl/-/xterm-addon-webgl-0.16.0.tgz",
+      "integrity": "sha512-E8cq1AiqNOv0M/FghPT+zPAEnvIQRDbAbkb04rRYSxUym69elPWVJ4sv22FCLBqM/3LcrmBLl/pELnBebVFKgA==",
+      "deprecated": "This package is now deprecated. Move to @xterm/addon-webgl instead.",
+      "license": "MIT",
+      "peerDependencies": {
+        "xterm": "^5.0.0"
+      }
+    },
     "node_modules/yallist": {
       "version": "3.1.1",
       "resolved": "https://registry.npmjs.org/yallist/-/yallist-3.1.1.tgz",

apps/web/package.json

@@ -22,7 +22,8 @@
     "tailwindcss": "^3.3.0",
     "xterm": "^5.3.0",
     "xterm-addon-fit": "^0.8.0",
-    "xterm-addon-web-links": "^0.9.0"
+    "xterm-addon-web-links": "^0.9.0",
+    "xterm-addon-webgl": "^0.16.0"
   },
   "devDependencies": {
     "@testing-library/jest-dom": "^6.9.1",

apps/web/src/api/projects.ts

@@ -2,50 +2,53 @@ import { apiClient } from "./client";
 import type { Project, ProjectWithRepos } from "../types";
 
 export type ProjectCreateInput = {
-  name: string;
-  description?: string | null;
+	name: string;
+	description?: string | null;
 };
 
 export type ProjectUpdateInput = {
-  name?: string | null;
-  description?: string | null;
+	name?: string | null;
+	description?: string | null;
 };
 
 export type SetDefaultSSHKeyInput = {
-  ssh_key_id: string;
+	ssh_key_id: string;
 };
 
 export const listProjects = async (): Promise<ProjectWithRepos[]> => {
-  const response = await apiClient.get<ProjectWithRepos[]>("/projects");
-  return response.data;
+	const response = await apiClient.get<ProjectWithRepos[]>("/projects");
+	return response.data;
 };
 
 export const createProject = async (
-  input: ProjectCreateInput
+	input: ProjectCreateInput,
 ): Promise<Project> => {
-  const response = await apiClient.post<Project>("/projects", input);
-  return response.data;
+	const response = await apiClient.post<Project>("/projects", input);
+	return response.data;
 };
 
 export const updateProject = async (
-  projectId: string,
-  input: ProjectUpdateInput
+	projectId: string,
+	input: ProjectUpdateInput,
 ): Promise<Project> => {
-  const response = await apiClient.patch<Project>(`/projects/${projectId}`, input);
-  return response.data;
+	const response = await apiClient.patch<Project>(
+		`/projects/${projectId}`,
+		input,
+	);
+	return response.data;
 };
 
 export const deleteProject = async (projectId: string): Promise<void> => {
-  await apiClient.delete(`/projects/${projectId}`);
+	await apiClient.delete(`/projects/${projectId}`);
 };
 
 export const setDefaultSSHKey = async (
-  projectId: string,
-  input: SetDefaultSSHKeyInput
+	projectId: string,
+	input: SetDefaultSSHKeyInput,
 ): Promise<Project> => {
-  const response = await apiClient.patch<Project>(
-    `/projects/${projectId}/default-ssh-key`,
-    input
-  );
-  return response.data;
+	const response = await apiClient.patch<Project>(
+		`/projects/${projectId}/default-ssh-key`,
+		input,
+	);
+	return response.data;
 };

apps/web/src/api/workspaces.ts

@@ -9,7 +9,7 @@ import type {
 
 function workspaceUrl(projectId: string, repoId: string, workspaceId?: string) {
 	const base = `/projects/${projectId}/repositories/${repoId}/workspaces`;
-	return workspaceId ? `${base}/${workspaceId}/` : `${base}/`;
+	return workspaceId ? `${base}/${workspaceId}` : `${base}/`;
 }
 
 export async function listWorkspaces(
@@ -39,6 +39,13 @@ export async function createWorkspace(
 	return response.data;
 }
 
+export async function createWorkspaceTopLevel(
+	data: CreateWorkspaceRequest & { repo_id: string },
+): Promise<Workspace> {
+	const response = await apiClient.post<Workspace>("/workspaces/", data);
+	return response.data;
+}
+
 export async function getWorkspace(
 	projectId: string,
 	repoId: string,
@@ -64,13 +71,11 @@ export async function updateWorkspace(
 }
 
 export async function deleteWorkspace(
-	projectId: string,
-	repoId: string,
 	workspaceId: string,
 	force = false,
 ): Promise<{ status: string }> {
 	const response = await apiClient.delete<{ status: string }>(
-		`${workspaceUrl(projectId, repoId, workspaceId)}?force=${force}`,
+		`/workspaces/${workspaceId}?force=${force}`,
 	);
 	return response.data;
 }

apps/web/src/components/app-shell.tsx

@@ -14,6 +14,7 @@ import { EventToastBridge } from "./event-toast-bridge";
 import { NotificationCenter } from "./notification-center";
 import { Icon } from "./icon";
 import { MobileNav } from "./mobile-nav";
+import { StartToolFAB } from "./start-tool-fab";
 import type { IconName } from "../utils/icons";
 
 const NAV_ITEMS: {
@@ -34,11 +35,23 @@ const NAV_ITEMS: {
 const SessionItem = ({ session }: { session: Session }) => {
 	const isRunning = session.status === "running";
 
+	// Determine the link target:
+	// - Web tools open their tunnel URL
+	// - Terminal tools open the terminal page
+	// - Everything else falls back to the project page
+	const hasTerminal = session.tool_type_interfaces.includes("terminal");
+	const hasWeb = session.tool_type_interfaces.includes("web");
+	const href = session.url && hasWeb
+		? session.url
+		: hasTerminal
+			? `/instances/${session.id}/terminal`
+			: `/projects/${session.project_id}`;
+
 	return (
 		<a
-			href={session.url ?? `/projects/${session.project_id}`}
-			target={session.url ? "_blank" : undefined}
-			rel={session.url ? "noopener noreferrer" : undefined}
+			href={href}
+			target="_blank"
+			rel="noopener noreferrer"
 			className="nav-item session-item"
 			title={`${session.display_name} (${session.status})`}
 		>
@@ -170,6 +183,7 @@ export const AppShell = () => {
 								}
 							/>
 						)}
+						<StartToolFAB />
 					</div>
 				</NotificationProvider>
 			</ToastProvider>

apps/web/src/components/session-card.tsx

@@ -58,6 +58,11 @@ export function SessionCard({
 	const isTerminalOnly =
 		session.tool_type_interfaces?.includes("terminal") &&
 		!session.tool_type_interfaces?.includes("web");
+	const openHref = session.url
+		? session.url
+		: isTerminalOnly
+			? `/instances/${session.id}/terminal`
+			: undefined;
 	const hasTunnelError =
 		!isTerminalOnly && tunnelHealth?.tunnel_status === "unreachable";
 	const hasAppError =
@@ -150,9 +155,9 @@ export function SessionCard({
 				<div className="session-card-actions mobile">
 					{isActive && (
 						<>
-							{session.url ? (
+							{openHref ? (
 								<a
-									href={session.url}
+									href={openHref}
 									target="_blank"
 									rel="noopener noreferrer"
 									className="secondary-button mobile-primary"
@@ -207,9 +212,9 @@ export function SessionCard({
 				<div className="session-card-actions">
 					{isActive && (
 						<>
-							{session.url ? (
+							{openHref ? (
 								<a
-									href={session.url}
+									href={openHref}
 									target="_blank"
 									rel="noopener noreferrer"
 									className="secondary-button small"

apps/web/src/components/start-tool-fab.tsx

@@ -0,0 +1,118 @@
+/** Floating action button to start a tool from any page. */
+
+import { useState } from "react";
+import { Icon } from "./icon";
+import { ToolStarter } from "./tool-starter";
+import type { Workspace } from "../types/workspace";
+import { listAllWorkspaces } from "../api/workspaces";
+
+export function StartToolFAB() {
+	const [open, setOpen] = useState(false);
+	const [workspaces, setWorkspaces] = useState<Workspace[]>([]);
+	const [workspacesLoading, setWorkspacesLoading] = useState(false);
+	const [selectedWorkspace, setSelectedWorkspace] = useState<Workspace | null>(
+		null,
+	);
+
+	const handleOpen = async () => {
+		setOpen(true);
+		setWorkspacesLoading(true);
+		try {
+			const data = await listAllWorkspaces();
+			setWorkspaces(data);
+		} catch {
+			// ignore
+		} finally {
+			setWorkspacesLoading(false);
+		}
+	};
+
+	const handleClose = () => {
+		setOpen(false);
+		setSelectedWorkspace(null);
+	};
+
+	return (
+		<>
+			<button
+				className="start-tool-fab"
+				onClick={handleOpen}
+				title="Start a new tool"
+				type="button"
+				aria-label="Start a new tool"
+			>
+				<Icon name="play" size="md" />
+			</button>
+
+			{open && (
+				<div className="modal-overlay" onClick={handleClose}>
+					<div
+						className="modal-content start-tool-modal"
+						onClick={(e) => e.stopPropagation()}
+					>
+						<div className="modal-header">
+							<h3>Start Tool</h3>
+							<button
+								className="ghost-button small"
+								onClick={handleClose}
+								type="button"
+								aria-label="Close"
+							>
+								<Icon name="close" size="sm" />
+							</button>
+						</div>
+
+						{workspacesLoading ? (
+							<p className="muted">Loading workspaces...</p>
+						) : workspaces.length === 0 ? (
+							<p className="muted">
+								No workspaces yet.{" "}
+								<a href="/workspaces">Create a workspace first</a>.
+							</p>
+						) : !selectedWorkspace ? (
+							<div className="form-group">
+								<label htmlFor="fab-workspace-select">Select a workspace</label>
+								<select
+									id="fab-workspace-select"
+									value=""
+									onChange={(e) => {
+										const ws = workspaces.find((w) => w.id === e.target.value);
+										if (ws) setSelectedWorkspace(ws);
+									}}
+								>
+									<option value="">Choose a workspace...</option>
+									{workspaces.map((ws) => (
+										<option key={ws.id} value={ws.id}>
+											{ws.project_name} / {ws.repo_name} / {ws.name}
+										</option>
+									))}
+								</select>
+							</div>
+						) : (
+							<>
+								<div className="tool-starter-header">
+									<h4>
+										{selectedWorkspace.project_name} /{" "}
+										{selectedWorkspace.repo_name} / {selectedWorkspace.name}
+									</h4>
+									<button
+										className="ghost-button small"
+										onClick={() => setSelectedWorkspace(null)}
+										type="button"
+									>
+										Change
+									</button>
+								</div>
+								<ToolStarter
+									workspace={selectedWorkspace}
+									onStarted={handleClose}
+									onCancel={() => setSelectedWorkspace(null)}
+								/>
+							</>
+						)}
+					</div>
+				</div>
+			)}
+		</>
+	);
+}

apps/web/src/components/start-tool-modal.tsx

@@ -2,6 +2,8 @@
 
 import { useState } from "react";
 import { Icon } from "./icon";
+import { listToolTypes, type ToolType } from "../api/tool_types";
+import { useAsyncData } from "../hooks/use-async-data";
 import type { Workspace } from "../types/workspace";
 
 export interface StartToolModalProps {
@@ -20,6 +22,12 @@ export function StartToolModal({
 	const [submitting, setSubmitting] = useState(false);
 	const [error, setError] = useState<string | null>(null);
 
+	const {
+		data: toolTypes,
+		status,
+		error: loadError,
+	} = useAsyncData<ToolType[]>(listToolTypes, []);
+
 	const handleSubmit = async (e: React.FormEvent) => {
 		e.preventDefault();
 		if (!toolTypeId) {
@@ -56,13 +64,19 @@ export function StartToolModal({
 							id="tool-type"
 							value={toolTypeId}
 							onChange={(e) => setToolTypeId(e.target.value)}
-							disabled={submitting}
+							disabled={submitting || status === "loading"}
 						>
 							<option value="">Select a tool...</option>
-							<option value="code-server">Code Server</option>
-							<option value="jupyter-notebook">Jupyter Notebook</option>
-							<option value="terminal">Terminal</option>
+							{toolTypes?.map((tt) => (
+								<option key={tt.id} value={tt.id}>
+									{tt.display_name}
+								</option>
+							))}
 						</select>
+						{status === "loading" && (
+							<span className="muted">Loading tools...</span>
+						)}
+						{loadError && <span className="error-text">{loadError}</span>}
 					</div>
 					<div className="form-group">
 						<label htmlFor="config-profile">Config Profile (optional)</label>
@@ -88,7 +102,7 @@ export function StartToolModal({
 						<button
 							type="submit"
 							className="btn btn-primary"
-							disabled={submitting}
+							disabled={submitting || status !== "ready"}
 						>
 							{submitting ? "Starting..." : "Start Tool"}
 						</button>

apps/web/src/components/terminal.tsx

@@ -8,6 +8,7 @@ import React, {
 import { Terminal } from "xterm";
 import { FitAddon } from "xterm-addon-fit";
 import { WebLinksAddon } from "xterm-addon-web-links";
+import { WebglAddon } from "xterm-addon-webgl";
 import "xterm/css/xterm.css";
 
 import {
@@ -107,6 +108,7 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 
 			// WebSocket connection established
 			const ws = new WebSocket(wsUrl);
+			ws.binaryType = "arraybuffer";
 			wsRef.current = ws;
 
 			ws.onopen = () => {
@@ -137,14 +139,35 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 				}, 30000);
 			};
 
+			// Flow control: accumulate processed bytes and send ack
+			let ackAccumulator = 0;
+			const ACK_THRESHOLD = 4096;
+			let ackTimeout: ReturnType<typeof setTimeout> | null = null;
+
+			const flushAck = () => {
+				if (ackAccumulator > 0 && ws.readyState === WebSocket.OPEN) {
+					ws.send(JSON.stringify({ type: "ack", chars: ackAccumulator }));
+					ackAccumulator = 0;
+				}
+			};
+
 			ws.onmessage = (event) => {
 				if (!termRef.current) return;
 
-				if (event.data instanceof Blob) {
-					event.data.arrayBuffer().then((buffer) => {
-						const data = new Uint8Array(buffer);
-						termRef.current?.write(data);
-					});
+				if (event.data instanceof ArrayBuffer) {
+					const data = new Uint8Array(event.data);
+					termRef.current.write(data);
+
+					// Flow control: accumulate processed bytes
+					ackAccumulator += data.length;
+					if (ackAccumulator >= ACK_THRESHOLD) {
+						flushAck();
+					} else if (!ackTimeout) {
+						ackTimeout = setTimeout(() => {
+							ackTimeout = null;
+							flushAck();
+						}, 100);
+					}
 				} else if (typeof event.data === "string") {
 					try {
 						const msg = JSON.parse(event.data);
@@ -251,6 +274,11 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 				lineHeight: 1.2,
 				letterSpacing: 0,
 				allowTransparency: false,
+				scrollback: 10000,
+				ignoreBracketedPasteMode: false,
+				fastScrollSensitivity: 5,
+				scrollSensitivity: 1,
+				smoothScrollDuration: 0,
 				theme: {
 					background: "#1e1e1e",
 					foreground: "#d4d4d4",
@@ -282,9 +310,31 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 			term.loadAddon(fitAddon);
 			term.loadAddon(new WebLinksAddon());
 
+			// Load WebGL renderer for GPU acceleration, fall back to DOM
+			let webglAddon: WebglAddon | null = null;
+			try {
+				webglAddon = new WebglAddon();
+				term.loadAddon(webglAddon);
+				webglAddon.onContextLoss(() => {
+					console.warn("WebGL context lost, falling back to DOM renderer");
+					try {
+						webglAddon?.dispose();
+					} catch {
+						// ignore
+					}
+					webglAddon = null;
+					// Trigger a refit since cell dimensions may differ
+					requestAnimationFrame(() => fitTerminal());
+				});
+			} catch (e) {
+				console.warn("WebGL renderer failed to load, using DOM renderer", e);
+			}
+
 			const container = terminalRef.current;
 
 			// Define fitTerminal before connectWebSocket so it's available in onmessage
+			let lastSentCols = 0;
+			let lastSentRows = 0;
 			const fitTerminal = () => {
 				if (!fitAddonRef.current || !termRef.current) return;
 				try {
@@ -294,18 +344,19 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 					return;
 				}
 				const { cols, rows } = termRef.current;
-				// Force refresh if dimensions are valid
-				if (cols > 0 && rows > 0) {
-					try {
-						termRef.current.refresh(0, rows - 1);
-					} catch {
-						// Ignore refresh errors
+				// Only send resize when dimensions actually changed
+				if (
+					cols > 0 &&
+					rows > 0 &&
+					(cols !== lastSentCols || rows !== lastSentRows)
+				) {
+					lastSentCols = cols;
+					lastSentRows = rows;
+					const currentWs = wsRef.current;
+					if (currentWs?.readyState === WebSocket.OPEN) {
+						currentWs.send(JSON.stringify({ type: "resize", cols, rows }));
 					}
 				}
-				const currentWs = wsRef.current;
-				if (currentWs?.readyState === WebSocket.OPEN && cols > 0 && rows > 0) {
-					currentWs.send(JSON.stringify({ type: "resize", cols, rows }));
-				}
 			};
 
 			// Open xterm first (must happen before fit)
@@ -352,16 +403,12 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 						// If the viewport is scrollable, scroll it directly.
 						// Otherwise we are in alternate screen (tmux/vim) and must
 						// send SGR 1006 mouse-wheel protocol data.
-						const hasScrollback =
-							viewport.scrollHeight > viewport.clientHeight;
+						const hasScrollback = viewport.scrollHeight > viewport.clientHeight;
 						if (hasScrollback) {
 							viewport.scrollTop += deltaY;
 						} else {
 							const ws = wsRef.current;
-							if (
-								ws?.readyState === WebSocket.OPEN &&
-								termRef.current
-							) {
+							if (ws?.readyState === WebSocket.OPEN && termRef.current) {
 								// Use the cursor position as the wheel location so
 								// tmux knows which pane to scroll.
 								const buf = termRef.current.buffer.active;
@@ -538,7 +585,21 @@ export const TerminalComponent = React.forwardRef<TerminalRef, TerminalProps>(
 					window.clearInterval(heartbeatCheckRef.current);
 					heartbeatCheckRef.current = null;
 				}
-				term.dispose();
+				// Dispose WebGL addon BEFORE the terminal to avoid race with
+				// RenderService.setRenderer accessing a disposed renderer
+				if (webglAddon) {
+					try {
+						webglAddon.dispose();
+					} catch {
+						// Ignore disposal errors from partially torn-down terminal
+					}
+					webglAddon = null;
+				}
+				try {
+					term.dispose();
+				} catch {
+					// Ignore disposal errors from partially torn-down terminal
+				}
 			};
 		}, [instanceId, connectWebSocket]);
 

apps/web/src/components/tool-starter.tsx

@@ -0,0 +1,299 @@
+/** Unified tool starter — workspace-first, fetches real tool types and config profiles. */
+
+import { useState, useEffect, useCallback } from "react";
+import { Icon } from "./icon";
+import { listToolTypes, type ToolType } from "../api/tool_types";
+import { listConfigProfiles, type ConfigProfile } from "../api/config_profiles";
+import { listSSHKeys, type SSHKey } from "../api/ssh_keys";
+import type { Workspace } from "../types/workspace";
+import type { ToolInstance } from "../api/sessions";
+
+export interface ToolStarterProps {
+	workspace: Workspace;
+	onStarted: (instance: ToolInstance) => void;
+	onCancel?: () => void;
+}
+
+export function ToolStarter({
+	workspace,
+	onStarted,
+	onCancel,
+}: ToolStarterProps) {
+	const [toolTypes, setToolTypes] = useState<ToolType[]>([]);
+	const [toolTypesLoading, setToolTypesLoading] = useState(true);
+	const [toolTypesError, setToolTypesError] = useState<string | null>(null);
+
+	const [selectedToolTypeId, setSelectedToolTypeId] = useState("");
+
+	const [profiles, setProfiles] = useState<ConfigProfile[]>([]);
+	const [profilesLoading, setProfilesLoading] = useState(false);
+	const [selectedProfileId, setSelectedProfileId] = useState("");
+
+	const [sshKeys, setSshKeys] = useState<SSHKey[]>([]);
+	const [sshKeysLoading, setSshKeysLoading] = useState(true);
+	const [selectedSshKeyIds, setSelectedSshKeyIds] = useState<string[]>([]);
+
+	const [starting, setStarting] = useState(false);
+	const [error, setError] = useState<string | null>(null);
+
+	// Fetch tool types on mount
+	useEffect(() => {
+		const load = async () => {
+			try {
+				const data = await listToolTypes();
+				setToolTypes(data);
+			} catch (err) {
+				setToolTypesError(
+					err instanceof Error ? err.message : "Failed to load tool types",
+				);
+			} finally {
+				setToolTypesLoading(false);
+			}
+		};
+		void load();
+	}, []);
+
+	// Fetch config profiles when tool type changes
+	useEffect(() => {
+		if (!selectedToolTypeId) {
+			setProfiles([]);
+			setSelectedProfileId("");
+			return;
+		}
+		const load = async () => {
+			setProfilesLoading(true);
+			try {
+				const data = await listConfigProfiles(
+					workspace.project_id,
+					selectedToolTypeId,
+				);
+				setProfiles(data);
+				// Auto-select default profile if available
+				const defaultProfile = data.find((p) => p.is_default);
+				if (defaultProfile) {
+					setSelectedProfileId(defaultProfile.id);
+				} else {
+					setSelectedProfileId("");
+				}
+			} catch {
+				setProfiles([]);
+			} finally {
+				setProfilesLoading(false);
+			}
+		};
+		void load();
+	}, [selectedToolTypeId, workspace.project_id]);
+
+	// Fetch SSH keys on mount
+	useEffect(() => {
+		const load = async () => {
+			try {
+				const data = await listSSHKeys();
+				setSshKeys(data);
+				// Auto-select the repository's SSH key if available
+				if (workspace.repo_ssh_key_id) {
+					setSelectedSshKeyIds([workspace.repo_ssh_key_id]);
+				}
+			} catch (err) {
+				console.error("Failed to load SSH keys:", err);
+			} finally {
+				setSshKeysLoading(false);
+			}
+		};
+		void load();
+	}, [workspace.repo_ssh_key_id]);
+
+	const repoHasSshKey = !!workspace.repo_ssh_key_id;
+	const repoSshKey = sshKeys.find((k) => k.id === workspace.repo_ssh_key_id);
+
+	const handleStart = useCallback(async () => {
+		if (!selectedToolTypeId) {
+			setError("Please select a tool type");
+			return;
+		}
+		setStarting(true);
+		setError(null);
+		try {
+			const { createInstance, startInstance } = await import("../api/sessions");
+			const instance = await createInstance(
+				workspace.project_id,
+				workspace.repo_id,
+				selectedToolTypeId,
+				workspace.name,
+				undefined,
+				undefined,
+				undefined,
+				selectedProfileId || undefined,
+				selectedSshKeyIds.length > 0 ? selectedSshKeyIds : undefined,
+				workspace.id,
+			);
+			await startInstance(
+				workspace.project_id,
+				workspace.repo_id,
+				instance.id,
+				selectedProfileId || undefined,
+				selectedSshKeyIds.length > 0 ? selectedSshKeyIds : undefined,
+			);
+			onStarted(instance);
+		} catch (err) {
+			setError(err instanceof Error ? err.message : "Failed to start tool");
+		} finally {
+			setStarting(false);
+		}
+	}, [selectedToolTypeId, selectedProfileId, workspace, onStarted]);
+
+	return (
+		<div className="tool-starter">
+			{/* Context header — read-only workspace info */}
+			<div className="tool-starter-context">
+				<div className="context-row">
+					<span className="context-label">Project</span>
+					<span className="context-value">{workspace.project_name}</span>
+				</div>
+				<div className="context-row">
+					<span className="context-label">Repository</span>
+					<span className="context-value">{workspace.repo_name}</span>
+				</div>
+				<div className="context-row">
+					<span className="context-label">Workspace</span>
+					<span className="context-value">{workspace.name}</span>
+					<span className="branch-badge">
+						<Icon name="branch" size="sm" /> {workspace.branch}
+					</span>
+				</div>
+			</div>
+
+			{/* Tool Type */}
+			<div className="form-group">
+				<label htmlFor="tool-type">Tool Type</label>
+				<select
+					id="tool-type"
+					value={selectedToolTypeId}
+					onChange={(e) => {
+						setSelectedToolTypeId(e.target.value);
+						setError(null);
+					}}
+					disabled={toolTypesLoading || starting}
+				>
+					<option value="">Select a tool...</option>
+					{toolTypes.map((tt) => (
+						<option key={tt.id} value={tt.id}>
+							{tt.display_name}
+							{tt.category && ` (${tt.category})`}
+						</option>
+					))}
+				</select>
+				{toolTypesLoading && <span className="muted">Loading tools...</span>}
+				{toolTypesError && <span className="error-text">{toolTypesError}</span>}
+			</div>
+
+			{/* Config Profile */}
+			{selectedToolTypeId && (
+				<div className="form-group">
+					<label htmlFor="config-profile">Config Profile</label>
+					<select
+						id="config-profile"
+						value={selectedProfileId}
+						onChange={(e) => setSelectedProfileId(e.target.value)}
+						disabled={profilesLoading || starting}
+					>
+						<option value="">Default (no profile)</option>
+						{profiles.map((p) => (
+							<option key={p.id} value={p.id}>
+								{p.name}
+								{p.is_default && " (default)"}
+							</option>
+						))}
+					</select>
+					{profilesLoading && (
+						<span className="muted">Loading profiles...</span>
+					)}
+					{profiles.length === 0 && !profilesLoading && (
+						<span className="muted">No custom profiles for this tool.</span>
+					)}
+				</div>
+			)}
+
+			{/* SSH Key Selection */}
+			<div className="form-group ssh-key-selection">
+				<label>SSH Keys</label>
+				{sshKeysLoading ? (
+					<span className="muted">Loading SSH keys...</span>
+				) : sshKeys.length === 0 ? (
+					<span className="muted">No SSH keys configured.</span>
+				) : (
+					<div style={{ display: "flex", flexWrap: "wrap", gap: "0.5rem" }}>
+						{sshKeys.map((key) => (
+							<label
+								key={key.id}
+								className="checkbox-label"
+								style={{
+									display: "flex",
+									alignItems: "center",
+									gap: "0.25rem",
+									padding: "0.375rem 0.75rem",
+									background: "var(--panel)",
+									borderRadius: "0.375rem",
+									border: "1px solid var(--border)",
+									cursor: "pointer",
+								}}
+							>
+								<input
+									type="checkbox"
+									checked={selectedSshKeyIds.includes(key.id)}
+									onChange={(e) => {
+										if (e.target.checked) {
+											setSelectedSshKeyIds((prev) => [...prev, key.id]);
+										} else {
+											setSelectedSshKeyIds((prev) =>
+												prev.filter((id) => id !== key.id),
+											);
+										}
+									}}
+									disabled={starting}
+								/>
+								{key.name}
+							</label>
+						))}
+					</div>
+				)}
+				{!sshKeysLoading && repoHasSshKey && repoSshKey && (
+					<div className="hint" style={{ marginTop: "0.5rem" }}>
+						Repository key <strong>{repoSshKey.name}</strong> is pre-selected.
+					</div>
+				)}
+			</div>
+
+			{error && <p className="form-error">{error}</p>}
+
+			<div className="form-actions">
+				{onCancel && (
+					<button
+						type="button"
+						className="btn btn-secondary"
+						onClick={onCancel}
+						disabled={starting}
+					>
+						Cancel
+					</button>
+				)}
+				<button
+					type="button"
+					className="btn btn-primary"
+					onClick={handleStart}
+					disabled={!selectedToolTypeId || toolTypesLoading || starting}
+				>
+					{starting ? (
+						<>
+							<Icon name="loading" size="sm" /> Starting...
+						</>
+					) : (
+						<>
+							<Icon name="play" size="sm" /> Start Tool
+						</>
+					)}
+				</button>
+			</div>
+		</div>
+	);
+}

apps/web/src/components/workspace-card.tsx

@@ -2,6 +2,7 @@
 
 import { Link } from "react-router-dom";
 import { Icon } from "./icon";
+import { WorkspaceInstanceChips } from "./workspace-instance-chips";
 import type { Workspace } from "../types/workspace";
 
 export interface WorkspaceCardProps {
@@ -28,7 +29,10 @@ export function WorkspaceCard({
 
 	return (
 		<article className={`card workspace-card ${loading ? "loading" : ""}`}>
-			<Link to={`/workspaces/${workspace.id}`} className="workspace-header-link">
+			<Link
+				to={`/workspaces/${workspace.id}`}
+				className="workspace-header-link"
+			>
 				<div className="workspace-header">
 					<h4>{workspace.name}</h4>
 					<span className={`status-badge ${statusClass}`}>
@@ -43,12 +47,7 @@ export function WorkspaceCard({
 				<p className="workspace-branch">
 					<Icon name="branch" size="sm" /> {workspace.branch}
 				</p>
-				{workspace.instance_count > 0 && (
-					<p className="workspace-instances">
-						{workspace.instance_count} active tool
-						{workspace.instance_count > 1 ? "s" : ""}
-					</p>
-				)}
+				<WorkspaceInstanceChips workspaceId={workspace.id} />
 			</div>
 			<div className="workspace-actions">
 				<button

apps/web/src/components/workspace-create-form.tsx

@@ -1,37 +1,149 @@
-/** Form for creating a new workspace. */
+/** Unified workspace creation form with project/repo/branch selectors. */
 
-import { useState } from "react";
+import { useState, useEffect, useCallback } from "react";
 import { Icon } from "./icon";
-import type { CreateWorkspaceRequest } from "../types/workspace";
+import { listProjects } from "../api/projects";
+import { listRepositories } from "../api/git_repositories";
+import { createWorkspaceTopLevel } from "../api/workspaces";
+import { useGitRepo } from "../hooks/use-git-repo";
+import type { ProjectWithRepos } from "../types";
+import type { GitRepository } from "../api/git_repositories";
 
 export interface WorkspaceCreateFormProps {
-	projectId: string;
-	repoId: string;
-	defaultBranch?: string;
-	onSubmit: (data: CreateWorkspaceRequest) => Promise<void>;
+	/** Called after successful creation. */
+	onSubmit: () => void | Promise<void>;
+	/** Cancel callback. */
 	onCancel: () => void;
+	/** Optional: pre-selected project ID (hides project selector). */
+	defaultProjectId?: string;
+	/** Optional: pre-selected repo ID (hides repo selector). */
+	defaultRepoId?: string;
 }
 
 export function WorkspaceCreateForm({
-	defaultBranch = "main",
 	onSubmit,
 	onCancel,
+	defaultProjectId,
+	defaultRepoId,
 }: WorkspaceCreateFormProps) {
+	const isContextual = Boolean(defaultProjectId && defaultRepoId);
+
+	const [projects, setProjects] = useState<ProjectWithRepos[]>([]);
+	const [repos, setRepos] = useState<GitRepository[]>([]);
+	const [selectedProject, setSelectedProject] = useState(
+		defaultProjectId ?? "",
+	);
+	const [selectedRepo, setSelectedRepo] = useState(defaultRepoId ?? "");
+	const [selectedBranch, setSelectedBranch] = useState("");
+	const [newBranchName, setNewBranchName] = useState("");
+	const [isNewBranch, setIsNewBranch] = useState(false);
 	const [name, setName] = useState("");
-	const [branch, setBranch] = useState(defaultBranch);
 	const [submitting, setSubmitting] = useState(false);
+	const [fetchingProjects, setFetchingProjects] = useState(!isContextual);
 	const [error, setError] = useState<string | null>(null);
 
+	/* Git repo hook handles branch fetching, loading, errors */
+	const git = useGitRepo(
+		selectedProject || undefined,
+		selectedRepo || undefined,
+	);
+
+	/* Sync local branch state with hook data */
+	useEffect(() => {
+		if (git.branches.length > 0 && !selectedBranch) {
+			const preferred =
+				git.defaultBranch && git.branches.includes(git.defaultBranch)
+					? git.defaultBranch
+					: git.branches[0];
+			setSelectedBranch(preferred);
+			setIsNewBranch(false);
+		} else if (git.error && git.branches.length === 0 && !isNewBranch) {
+			// API failed — default to manual entry so user can type a branch
+			setIsNewBranch(true);
+			setSelectedBranch("__manual__");
+		}
+	}, [git.branches, git.defaultBranch, git.error, selectedBranch, isNewBranch]);
+
+	/* ── Load projects (standalone mode only) ── */
+	const loadProjects = useCallback(async () => {
+		if (isContextual) return;
+		try {
+			const data = await listProjects();
+			setProjects(data);
+			if (data.length === 1 && !defaultProjectId) {
+				setSelectedProject(data[0].id);
+			}
+		} catch {
+			setError("Failed to load projects");
+		} finally {
+			setFetchingProjects(false);
+		}
+	}, [isContextual, defaultProjectId]);
+
+	useEffect(() => {
+		void loadProjects();
+	}, [loadProjects]);
+
+	/* ── Load repos when project changes ── */
+	useEffect(() => {
+		if (!selectedProject) {
+			setRepos([]);
+			if (!defaultRepoId) setSelectedRepo("");
+			return;
+		}
+		const loadRepos = async () => {
+			try {
+				const data = await listRepositories(selectedProject);
+				setRepos(data);
+				if (data.length === 1 && !defaultRepoId) {
+					setSelectedRepo(data[0].id);
+				}
+			} catch {
+				setError("Failed to load repositories");
+			}
+		};
+		void loadRepos();
+	}, [selectedProject, defaultRepoId]);
+
+	const handleBranchChange = (value: string) => {
+		if (value === "__new__") {
+			setIsNewBranch(true);
+			setSelectedBranch("__new__");
+			setNewBranchName("");
+		} else if (value === "__manual__") {
+			setIsNewBranch(true);
+			setSelectedBranch("__manual__");
+			setNewBranchName("");
+		} else {
+			setIsNewBranch(false);
+			setSelectedBranch(value);
+		}
+	};
+
 	const handleSubmit = async (e: React.FormEvent) => {
 		e.preventDefault();
+		if (!selectedRepo) {
+			setError("Please select a repository");
+			return;
+		}
 		if (!name.trim()) {
 			setError("Workspace name is required");
 			return;
 		}
+		const branchName = isNewBranch ? newBranchName.trim() : selectedBranch;
+		if (!branchName) {
+			setError("Please select or enter a branch");
+			return;
+		}
 		setSubmitting(true);
 		setError(null);
 		try {
-			await onSubmit({ name: name.trim(), branch: branch.trim() });
+			await createWorkspaceTopLevel({
+				repo_id: selectedRepo,
+				name: name.trim(),
+				branch: branchName,
+			});
+			await onSubmit();
 		} catch (err) {
 			setError(
 				err instanceof Error ? err.message : "Failed to create workspace",
@@ -41,49 +153,169 @@ export function WorkspaceCreateForm({
 		}
 	};
 
+	/* Show single combined error */
+	const displayError = error || git.error;
+
+	if (fetchingProjects) {
+		return (
+			<div className="card workspace-create-inline">
+				<p className="muted">Loading projects...</p>
+			</div>
+		);
+	}
+
+	const branchSelectDisabled =
+		!selectedRepo || submitting || (git.loading && git.branches.length === 0);
+
 	return (
-		<form className="workspace-create-form card" onSubmit={handleSubmit}>
+		<div className="card workspace-create-inline">
 			<h3>
 				<Icon name="add" size="sm" /> Create Workspace
 			</h3>
-			<div className="form-group">
-				<label htmlFor="ws-name">Name</label>
-				<input
-					id="ws-name"
-					type="text"
-					value={name}
-					onChange={(e) => setName(e.target.value)}
-					placeholder="e.g., feature-branch"
-					disabled={submitting}
-				/>
-			</div>
-			<div className="form-group">
-				<label htmlFor="ws-branch">
-					<Icon name="branch" size="sm" /> Branch
-				</label>
-				<input
-					id="ws-branch"
-					type="text"
-					value={branch}
-					onChange={(e) => setBranch(e.target.value)}
-					placeholder="main"
-					disabled={submitting}
-				/>
-			</div>
-			{error && <p className="form-error">{error}</p>}
-			<div className="form-actions">
-				<button
-					type="button"
-					className="btn btn-secondary"
-					onClick={onCancel}
-					disabled={submitting}
-				>
-					Cancel
-				</button>
-				<button type="submit" className="btn btn-primary" disabled={submitting}>
-					{submitting ? "Creating..." : "Create"}
-				</button>
-			</div>
-		</form>
+			<form onSubmit={handleSubmit} className="workspace-create-form-grid">
+				{/* Project selector (standalone only) */}
+				{!isContextual && (
+					<div className="form-group">
+						<label>Project</label>
+						<select
+							value={selectedProject}
+							onChange={(e) => {
+								setSelectedProject(e.target.value);
+								setSelectedBranch("");
+							}}
+							required
+						>
+							<option value="">Select project...</option>
+							{projects.map((p) => (
+								<option key={p.id} value={p.id}>
+									{p.name}
+								</option>
+							))}
+						</select>
+					</div>
+				)}
+
+				{/* Repo selector (standalone only) */}
+				{!isContextual && (
+					<div className="form-group">
+						<label>Repository</label>
+						<select
+							value={selectedRepo}
+							onChange={(e) => {
+								setSelectedRepo(e.target.value);
+								setSelectedBranch("");
+							}}
+							required
+							disabled={!selectedProject || repos.length === 0}
+						>
+							<option value="">
+								{!selectedProject
+									? "Select a project first"
+									: repos.length === 0
+										? "No repositories"
+										: "Select repository..."}
+							</option>
+							{repos.map((r) => (
+								<option key={r.id} value={r.id}>
+									{r.name}
+								</option>
+							))}
+						</select>
+					</div>
+				)}
+
+				<div className="form-group">
+					<label>Name</label>
+					<input
+						type="text"
+						value={name}
+						onChange={(e) => setName(e.target.value)}
+						placeholder="e.g., feature-branch"
+						required
+						disabled={submitting}
+					/>
+				</div>
+
+				<div className="form-group">
+					<label>
+						<Icon name="branch" size="sm" /> Branch
+					</label>
+
+					{/* Show a hint when branches couldn’t be loaded */}
+					{git.error && git.branches.length === 0 && selectedRepo && (
+						<p
+							className="muted"
+							style={{
+								fontSize: "var(--font-size-xs)",
+								marginBottom: "0.25rem",
+							}}
+						>
+							Couldn’t load branches — type one manually.
+						</p>
+					)}
+
+					<select
+						value={selectedBranch}
+						onChange={(e) => handleBranchChange(e.target.value)}
+						required
+						disabled={branchSelectDisabled}
+					>
+						<option value="">
+							{git.loading && git.branches.length === 0
+								? "Loading branches..."
+								: !selectedRepo
+									? "Select a repository first"
+									: "Select branch..."}
+						</option>
+
+						{git.branches.map((b) => (
+							<option key={b} value={b}>
+								{b}
+								{b === git.defaultBranch ? " (default)" : ""}
+							</option>
+						))}
+
+						<option value="__new__">+ Create new branch...</option>
+					</select>
+
+					{/* Text input for new branch or manual entry */}
+					{isNewBranch && (
+						<input
+							type="text"
+							value={newBranchName}
+							onChange={(e) => setNewBranchName(e.target.value)}
+							placeholder="new-branch-name"
+							required
+							style={{ marginTop: "0.5rem" }}
+							disabled={submitting}
+						/>
+					)}
+				</div>
+
+				{displayError && (
+					<div className="form-error" style={{ gridColumn: "1 / -1" }}>
+						{displayError}
+					</div>
+				)}
+
+				<div className="form-actions" style={{ gridColumn: "1 / -1" }}>
+					<button
+						type="button"
+						className="btn btn-secondary"
+						onClick={onCancel}
+						disabled={submitting}
+					>
+						Cancel
+					</button>
+					<button
+						type="submit"
+						className="btn btn-primary"
+						disabled={submitting || !selectedRepo}
+					>
+						{submitting ? "Creating..." : "Create Workspace"}
+					</button>
+				</div>
+			</form>
+		</div>
 	);
 }

apps/web/src/components/workspace-instance-chips.tsx

@@ -0,0 +1,57 @@
+/** Small component showing running instances for a workspace. */
+
+import { useEffect, useState } from "react";
+import { listWorkspaceInstances } from "../api/workspace-instances";
+import type { ToolInstance } from "../api/sessions";
+
+interface WorkspaceInstanceChipsProps {
+	workspaceId: string;
+}
+
+export function WorkspaceInstanceChips({
+	workspaceId,
+}: WorkspaceInstanceChipsProps) {
+	const [instances, setInstances] = useState<ToolInstance[]>([]);
+	const [loading, setLoading] = useState(true);
+
+	useEffect(() => {
+		const load = async () => {
+			try {
+				const data = await listWorkspaceInstances(workspaceId);
+				setInstances(data);
+			} catch {
+				// ignore
+			} finally {
+				setLoading(false);
+			}
+		};
+		void load();
+	}, [workspaceId]);
+
+	if (loading) return <span className="muted">...</span>;
+	if (instances.length === 0) return null;
+
+	return (
+		<div className="instance-chips">
+			{instances.map((inst) => (
+				<span
+					key={inst.id}
+					className={`instance-chip ${inst.status}`}
+					title={inst.display_name}
+				>
+					{inst.display_name}
+					{inst.status === "running" && inst.url && (
+						<a
+							href={inst.url}
+							target="_blank"
+							rel="noopener noreferrer"
+							onClick={(e) => e.stopPropagation()}
+						>
+							↗
+						</a>
+					)}
+				</span>
+			))}
+		</div>
+	);
+}

apps/web/src/hooks/use-git-repo.ts

@@ -0,0 +1,280 @@
+/** Unified hook for git repository operations.
+ *
+ * Centralizes branch fetching, status, history, and git actions
+ * so components don't duplicate this logic.
+ */
+
+import { useState, useEffect, useCallback } from "react";
+import {
+	listRepositoryBranches,
+	getRepositoryStatus,
+	getRepositoryHistory,
+	getCommitDetail,
+	commitChanges,
+	pushRepository,
+	pullRepository,
+	fetchRepository,
+	checkoutBranch,
+	createBranch,
+	deleteBranch,
+	mergeBranches,
+	type Branch,
+	type CommitHistoryResponse,
+	type CommitDetail,
+} from "../api/git_repositories";
+
+export interface GitStatus {
+	branch: string;
+	modified: string[];
+	added: string[];
+	deleted: string[];
+	untracked: string[];
+	renamed: string[];
+	ahead: number;
+	behind: number;
+}
+
+export interface UseGitRepoResult {
+	/** Available branch names. */
+	branches: string[];
+	/** The repo's default branch. */
+	defaultBranch: string;
+	/** Current working-directory status. */
+	status: GitStatus | null;
+	/** Commit history. */
+	history: CommitHistoryResponse | null;
+	/** Selected commit detail. */
+	commitDetail: CommitDetail | null;
+	/** True while any async operation is in flight. */
+	loading: boolean;
+	/** Error message from the last failed operation. */
+	error: string | null;
+	/** Refresh branches list. */
+	refreshBranches: () => Promise<void>;
+	/** Refresh working-directory status. */
+	refreshStatus: () => Promise<void>;
+	/** Refresh commit history. */
+	refreshHistory: (branch?: string, limit?: number) => Promise<void>;
+	/** Fetch a single commit's details. */
+	loadCommitDetail: (hash: string) => Promise<void>;
+	/** Stage + commit changes. */
+	commit: (message: string, files?: string[]) => Promise<void>;
+	/** Push current branch (or named branch) to remote. */
+	push: (branch?: string) => Promise<void>;
+	/** Pull from remote. */
+	pull: (branch?: string) => Promise<void>;
+	/** Fetch from remote. */
+	fetch: () => Promise<void>;
+	/** Checkout an existing branch. */
+	checkout: (branch: string) => Promise<void>;
+	/** Create and checkout a new branch. */
+	createBranch: (name: string, baseBranch?: string) => Promise<void>;
+	/** Delete a branch. */
+	deleteBranch: (name: string, force?: boolean) => Promise<void>;
+	/** Merge source into current (or target) branch. */
+	merge: (
+		sourceBranch: string,
+		targetBranch?: string,
+		message?: string,
+	) => Promise<void>;
+	/** Clear the current error. */
+	clearError: () => void;
+}
+
+export function useGitRepo(
+	projectId: string | undefined,
+	repoId: string | undefined,
+): UseGitRepoResult {
+	const [branches, setBranches] = useState<string[]>([]);
+	const [defaultBranch, setDefaultBranch] = useState("");
+	const [status, setStatus] = useState<GitStatus | null>(null);
+	const [history, setHistory] = useState<CommitHistoryResponse | null>(null);
+	const [commitDetail, setCommitDetail] = useState<CommitDetail | null>(null);
+	const [loading, setLoading] = useState(false);
+	const [error, setError] = useState<string | null>(null);
+
+	const extractError = (err: unknown): string => {
+		if (typeof err === "object" && err !== null) {
+			const e = err as Record<string, unknown>;
+			const response = e.response as Record<string, unknown> | undefined;
+			const data = response?.data as Record<string, unknown> | undefined;
+			if (typeof data?.detail === "string") return data.detail;
+			if (typeof data?.message === "string") return data.message;
+			if (typeof e.message === "string") return e.message;
+		}
+		return "Git operation failed";
+	};
+
+	const withLoading = useCallback(
+		async <T>(fn: () => Promise<T>): Promise<T> => {
+			setLoading(true);
+			setError(null);
+			try {
+				return await fn();
+			} catch (err) {
+				setError(extractError(err));
+				throw err;
+			} finally {
+				setLoading(false);
+			}
+		},
+		[],
+	);
+
+	const refreshBranches = useCallback(async () => {
+		if (!projectId || !repoId) return;
+		const data = await withLoading(() =>
+			listRepositoryBranches(projectId, repoId),
+		);
+		setBranches(data.branches.map((b: Branch) => b.name));
+		setDefaultBranch(data.default_branch ?? "");
+	}, [projectId, repoId, withLoading]);
+
+	const refreshStatus = useCallback(async () => {
+		if (!projectId || !repoId) return;
+		const data = await withLoading(() =>
+			getRepositoryStatus(projectId, repoId),
+		);
+		setStatus({
+			branch: data.branch,
+			modified: data.modified,
+			added: data.added,
+			deleted: data.deleted,
+			untracked: data.untracked,
+			renamed: data.renamed ?? [],
+			ahead: data.ahead,
+			behind: data.behind,
+		});
+	}, [projectId, repoId, withLoading]);
+
+	const refreshHistory = useCallback(
+		async (branch?: string, limit = 50) => {
+			if (!projectId || !repoId) return;
+			const data = await withLoading(() =>
+				getRepositoryHistory(projectId, repoId, branch, limit),
+			);
+			setHistory(data);
+		},
+		[projectId, repoId, withLoading],
+	);
+
+	const loadCommitDetail = useCallback(
+		async (hash: string) => {
+			if (!projectId || !repoId) return;
+			const data = await withLoading(() =>
+				getCommitDetail(projectId, repoId, hash),
+			);
+			setCommitDetail(data);
+		},
+		[projectId, repoId, withLoading],
+	);
+
+	const commit = useCallback(
+		async (message: string, files?: string[]) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() => commitChanges(projectId, repoId, message, files));
+			await refreshStatus();
+		},
+		[projectId, repoId, withLoading, refreshStatus],
+	);
+
+	const push = useCallback(
+		async (branch?: string) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() => pushRepository(projectId, repoId, branch));
+			await refreshStatus();
+		},
+		[projectId, repoId, withLoading, refreshStatus],
+	);
+
+	const pull = useCallback(
+		async (branch?: string) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() => pullRepository(projectId, repoId, branch));
+			await refreshStatus();
+		},
+		[projectId, repoId, withLoading, refreshStatus],
+	);
+
+	const fetch = useCallback(async () => {
+		if (!projectId || !repoId) return;
+		await withLoading(() => fetchRepository(projectId, repoId));
+		await refreshStatus();
+	}, [projectId, repoId, withLoading, refreshStatus]);
+
+	const checkout = useCallback(
+		async (branch: string) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() => checkoutBranch(projectId, repoId, branch));
+			await refreshStatus();
+			await refreshBranches();
+		},
+		[projectId, repoId, withLoading, refreshStatus, refreshBranches],
+	);
+
+	const createBranchFn = useCallback(
+		async (name: string, baseBranch = "HEAD") => {
+			if (!projectId || !repoId) return;
+			await withLoading(() =>
+				createBranch(projectId, repoId, name, baseBranch),
+			);
+			await refreshBranches();
+			await refreshStatus();
+		},
+		[projectId, repoId, withLoading, refreshBranches, refreshStatus],
+	);
+
+	const deleteBranchFn = useCallback(
+		async (name: string, force = false) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() => deleteBranch(projectId, repoId, name, force));
+			await refreshBranches();
+		},
+		[projectId, repoId, withLoading, refreshBranches],
+	);
+
+	const merge = useCallback(
+		async (sourceBranch: string, targetBranch?: string, message?: string) => {
+			if (!projectId || !repoId) return;
+			await withLoading(() =>
+				mergeBranches(projectId, repoId, sourceBranch, targetBranch, message),
+			);
+			await refreshStatus();
+			await refreshHistory();
+		},
+		[projectId, repoId, withLoading, refreshStatus, refreshHistory],
+	);
+
+	// Auto-refresh branches when projectId/repoId become valid
+	useEffect(() => {
+		if (projectId && repoId) {
+			void refreshBranches();
+		} else {
+			setBranches([]);
+			setDefaultBranch("");
+		}
+	}, [projectId, repoId, refreshBranches]);
+
+	return {
+		branches,
+		defaultBranch,
+		status,
+		history,
+		commitDetail,
+		loading,
+		error,
+		refreshBranches,
+		refreshStatus,
+		refreshHistory,
+		loadCommitDetail,
+		commit,
+		push,
+		pull,
+		fetch,
+		checkout,
+		createBranch: createBranchFn,
+		deleteBranch: deleteBranchFn,
+		merge,
+		clearError: () => setError(null),
+	};
+}

apps/web/src/hooks/use-instance-actions.ts

@@ -40,10 +40,10 @@ export function useInstanceActions(
 			return;
 		}
 		if (session.tool_type_interfaces?.includes("terminal")) {
-			window.location.href = `/instances/${session.id}/terminal`;
+			window.open(`/instances/${session.id}/terminal`, "_blank", "noopener,noreferrer");
 			return;
 		}
-		window.location.href = `/projects/${session.project_id}`;
+		window.open(`/projects/${session.project_id}`, "_blank", "noopener,noreferrer");
 	}, []);
 
 	const handleStart = useCallback(

apps/web/src/hooks/use-start-tool.ts

@@ -0,0 +1,64 @@
+/** Shared hook for starting a tool on a workspace. */
+
+import { useCallback, useState } from "react";
+import { createInstance, startInstance } from "../api/sessions";
+import type { Workspace } from "../types/workspace";
+import type { ToolInstance } from "../api/sessions";
+
+export interface UseStartToolResult {
+	starting: boolean;
+	error: string | null;
+	startTool: (
+		workspace: Workspace,
+		toolTypeId: string,
+		displayName?: string,
+		configProfileId?: string,
+	) => Promise<ToolInstance | null>;
+}
+
+export function useStartTool(): UseStartToolResult {
+	const [starting, setStarting] = useState(false);
+	const [error, setError] = useState<string | null>(null);
+
+	const startTool = useCallback(
+		async (
+			workspace: Workspace,
+			toolTypeId: string,
+			displayName?: string,
+			configProfileId?: string,
+		): Promise<ToolInstance | null> => {
+			setStarting(true);
+			setError(null);
+			try {
+				const instance = await createInstance(
+					workspace.project_id,
+					workspace.repo_id,
+					toolTypeId,
+					displayName || workspace.name,
+					undefined,
+					undefined,
+					undefined,
+					configProfileId,
+					[],
+					workspace.id,
+				);
+				await startInstance(
+					workspace.project_id,
+					workspace.repo_id,
+					instance.id,
+					configProfileId,
+				);
+				return instance;
+			} catch (err) {
+				const msg = err instanceof Error ? err.message : "Failed to start tool";
+				setError(msg);
+				return null;
+			} finally {
+				setStarting(false);
+			}
+		},
+		[],
+	);
+
+	return { starting, error, startTool };
+}

apps/web/src/hooks/use-workspace-actions.ts

@@ -17,8 +17,6 @@ export interface UseWorkspaceActionsResult {
 		data: CreateWorkspaceRequest,
 	) => Promise<Workspace>;
 	delete: (
-		projectId: string,
-		repoId: string,
 		workspace: Workspace,
 		onRefresh: () => Promise<void>,
 	) => Promise<void>;
@@ -60,15 +58,10 @@ export function useWorkspaceActions(): UseWorkspaceActionsResult {
 	);
 
 	const deleteAction = useCallback(
-		async (
-			projectId: string,
-			repoId: string,
-			workspace: Workspace,
-			onRefresh: () => Promise<void>,
-		) => {
+		async (workspace: Workspace, onRefresh: () => Promise<void>) => {
 			setLoadingId(workspace.id);
 			try {
-				await deleteWorkspace(projectId, repoId, workspace.id);
+				await deleteWorkspace(workspace.id);
 				await onRefresh();
 			} catch (err) {
 				const error = err as ApiError;
@@ -81,7 +74,7 @@ export function useWorkspaceActions(): UseWorkspaceActionsResult {
 							`\n\nDelete workspace and all instances?`,
 					);
 					if (confirmed) {
-						await deleteWorkspace(projectId, repoId, workspace.id, true);
+						await deleteWorkspace(workspace.id, true);
 						await onRefresh();
 					}
 				} else {
@@ -118,7 +111,7 @@ export function useWorkspaceActions(): UseWorkspaceActionsResult {
 						`${message}\n\nDelete this workspace?`,
 					);
 					if (confirmed) {
-						await deleteWorkspace(projectId, repoId, workspace.id, true);
+						await deleteWorkspace(workspace.id, true);
 						await onRefresh();
 					}
 				} else {

apps/web/src/pages/config-profiles.tsx

@@ -1539,7 +1539,6 @@ export const ConfigProfilesPage = () => {
 									onChange={(git_mounts) =>
 										updateFormField("git_mounts", git_mounts)
 									}
-	
 								/>
 							</div>
 

apps/web/src/pages/dashboard.tsx

@@ -2,265 +2,253 @@ import { useCallback, useEffect, useMemo, useState } from "react";
 import { useNavigate } from "react-router-dom";
 
 import { getDashboardSummary, type DashboardSummary } from "../api/dashboard";
-import { getUserSessions, checkInstanceHealth, type Session as SessionApi, type InstanceHealth } from "../api/sessions";
-import { listProjects } from "../api/projects";
-import { listRepositories, type GitRepository } from "../api/git_repositories";
-import { listToolTypes, type ToolType } from "../api/tool_types";
-import { updateUserConfig } from "../api/settings";
-import type { ProjectWithRepos } from "../types";
-import { EmptyState, ErrorState, LoadingState } from "../components/data-states";
-import { CreateSessionForm } from "../components/create-session-form";
+import {
+	getUserSessions,
+	checkInstanceHealth,
+	type Session as SessionApi,
+	type InstanceHealth,
+} from "../api/sessions";
+import { ErrorState, LoadingState } from "../components/data-states";
 import { SessionList } from "../components/session-list";
 import { useInstanceActions } from "../hooks/use-instance-actions";
 
 type HomeStatus = "loading" | "ready" | "error";
 
 const summaryCards = [
-  { label: "Open sessions", key: "openSessions" },
-  { label: "Projects", key: "projects" },
-  { label: "Repositories", key: "repositories" },
+	{ label: "Open sessions", key: "openSessions" },
+	{ label: "Projects", key: "projects" },
+	{ label: "Repositories", key: "repositories" },
 ] as const;
 
 type SessionView = SessionApi;
 
 export const HomePage = () => {
-  const navigate = useNavigate();
-  const [status, setStatus] = useState<HomeStatus>("loading");
-  const [summary, setSummary] = useState<DashboardSummary | null>(null);
-  const [sessions, setSessions] = useState<SessionView[]>([]);
-  const [projects, setProjects] = useState<ProjectWithRepos[]>([]);
-  const [repositories, setRepositories] = useState<GitRepository[]>([]);
-  const [toolTypes, setToolTypes] = useState<ToolType[]>([]);
-  const [selectedProject, setSelectedProject] = useState("");
-  const [tunnelHealth, setTunnelHealth] = useState<Record<string, InstanceHealth>>({});
-  const safeSessions = Array.isArray(sessions) ? sessions : [];
+	const navigate = useNavigate();
+	const [status, setStatus] = useState<HomeStatus>("loading");
+	const [summary, setSummary] = useState<DashboardSummary | null>(null);
+	const [sessions, setSessions] = useState<SessionView[]>([]);
+	const [tunnelHealth, setTunnelHealth] = useState<
+		Record<string, InstanceHealth>
+	>({});
+	const safeSessions = Array.isArray(sessions) ? sessions : [];
 
-  const loadHome = useCallback(async () => {
-    setStatus("loading");
-    try {
-      const [dashboard, sessionData, projectData, toolTypeData] = await Promise.all([
-        getDashboardSummary(),
-        getUserSessions(),
-        listProjects(),
-        listToolTypes(),
-      ]);
-      setSummary(dashboard);
-      setSessions(sessionData as SessionView[]);
-      setProjects(projectData);
-      setToolTypes(toolTypeData);
-      setStatus("ready");
-    } catch {
-      setStatus("error");
-    }
-  }, []);
+	const loadHome = useCallback(async () => {
+		setStatus("loading");
+		try {
+			const [dashboard, sessionData] = await Promise.all([
+				getDashboardSummary(),
+				getUserSessions(),
+			]);
+			setSummary(dashboard);
+			setSessions(sessionData as SessionView[]);
+			setStatus("ready");
+		} catch {
+			setStatus("error");
+		}
+	}, []);
 
-  useEffect(() => {
-    void loadHome();
-  }, [loadHome]);
+	useEffect(() => {
+		void loadHome();
+	}, [loadHome]);
 
-  const {
-    loadingSessionId: actionBusy,
-    handleOpen,
-    handleStart,
-    handleStop,
-    handleDelete,
-    handleRecreateTunnel,
-  } = useInstanceActions({ onRefresh: loadHome });
+	const {
+		loadingSessionId: actionBusy,
+		handleOpen,
+		handleStart,
+		handleStop,
+		handleDelete,
+		handleRecreateTunnel,
+	} = useInstanceActions({ onRefresh: loadHome });
 
-  // Poll tunnel health every 30 seconds for running instances
-  useEffect(() => {
-    const checkHealth = async () => {
-      const runningSessions = safeSessions.filter(
-        (s) => s.status === "running" && s.url
-      );
-      for (const session of runningSessions) {
-        try {
-          const health = await checkInstanceHealth(
-            session.project_id,
-            session.repository_id,
-            session.id
-          );
-          setTunnelHealth((prev) => ({
-            ...prev,
-            [session.id]: health,
-          }));
-        } catch {
-          setTunnelHealth((prev) => ({
-            ...prev,
-            [session.id]: {
-              healthy: false,
-              container_status: "unknown",
-              container_health: null,
-              container_exit_code: null,
-              tunnel_status: "error",
-              tunnel_status_code: null,
-              probe_status: "error",
-              last_probe_output: null,
-              error: "check failed",
-            },
-          }));
-        }
-      }
-    };
+	// Poll tunnel health every 30 seconds for running instances
+	useEffect(() => {
+		const checkHealth = async () => {
+			const runningSessions = safeSessions.filter(
+				(s) => s.status === "running" && s.url,
+			);
+			for (const session of runningSessions) {
+				try {
+					const health = await checkInstanceHealth(
+						session.project_id,
+						session.repository_id,
+						session.id,
+					);
+					setTunnelHealth((prev) => ({
+						...prev,
+						[session.id]: health,
+					}));
+				} catch {
+					setTunnelHealth((prev) => ({
+						...prev,
+						[session.id]: {
+							healthy: false,
+							container_status: "unknown",
+							container_health: null,
+							container_exit_code: null,
+							tunnel_status: "error",
+							tunnel_status_code: null,
+							probe_status: "error",
+							last_probe_output: null,
+							error: "check failed",
+						},
+					}));
+				}
+			}
+		};
 
-    void checkHealth();
-    const interval = setInterval(() => {
-      void checkHealth();
-    }, 30000);
-    return () => clearInterval(interval);
-  }, [safeSessions]);
+		void checkHealth();
+		const interval = setInterval(() => {
+			void checkHealth();
+		}, 30000);
+		return () => clearInterval(interval);
+	}, [safeSessions]);
 
-  useEffect(() => {
-    if (!selectedProject) {
-      setRepositories([]);
-      return;
-    }
+	const activeSessions = useMemo(
+		() =>
+			safeSessions.filter((session) =>
+				["running", "building", "pending"].includes(session.status),
+			),
+		[safeSessions],
+	);
 
-    const loadRepos = async () => {
-      try {
-        const data = await listRepositories(selectedProject);
-        setRepositories(data);
-      } catch {
-        setRepositories([]);
-      }
-    };
+	return (
+		<section className="stack home-page">
+			<header className="home-hero card">
+				<div className="stack-sm">
+					<p className="eyebrow">Workspace overview</p>
+					<h1>Home</h1>
+					<p className="muted">
+						Open sessions, available projects, and the fastest path back into
+						work.
+					</p>
+				</div>
+				<div className="home-hero-actions">
+					<button
+						className="primary-button"
+						type="button"
+						onClick={() => navigate("/projects")}
+					>
+						New Project
+					</button>
+					<button
+						className="secondary-button"
+						type="button"
+						onClick={() => navigate("/settings")}
+					>
+						Settings
+					</button>
+				</div>
+			</header>
 
-    void loadRepos();
-  }, [selectedProject]);
+			{status === "loading" && <LoadingState message="Loading overview..." />}
 
-  const activeSessions = useMemo(
-    () => safeSessions.filter((session) => ["running", "building", "pending"].includes(session.status)),
-    [safeSessions]
-  );
+			{status === "error" && (
+				<ErrorState
+					message="Unable to load your workspace overview."
+					onRetry={() => void loadHome()}
+				/>
+			)}
 
-  const handleCreateSuccess = async (instance: { id: string }) => {
-    await updateUserConfig({ last_session_id: instance.id });
-    setSelectedProject("");
-    await loadHome();
-  };
+			{status === "ready" && summary && (
+				<>
+					<div className="home-summary-grid">
+						{summaryCards.map((card) => (
+							<article className="card home-summary-card" key={card.label}>
+								<p className="card-label">{card.label}</p>
+								<p className="card-value">
+									{card.key === "openSessions"
+										? activeSessions.length
+										: card.key === "projects"
+											? summary.projects
+											: summary.repositories}
+								</p>
+							</article>
+						))}
+					</div>
 
-  return (
-    <section className="stack home-page">
-      <header className="home-hero card">
-        <div className="stack-sm">
-          <p className="eyebrow">Workspace overview</p>
-          <h1>Home</h1>
-          <p className="muted">Open sessions, available projects, and the fastest path back into work.</p>
-        </div>
-        <div className="home-hero-actions">
-          <button className="primary-button" type="button" onClick={() => navigate("/projects")}>New Project</button>
-          <button className="secondary-button" type="button" onClick={() => navigate("/settings")}>Settings</button>
-        </div>
-      </header>
+					<section className="card stack home-section">
+						<div className="page-header">
+							<div>
+								<p className="eyebrow">Open sessions</p>
+								<h2>
+									{
+										safeSessions.filter((s) =>
+											[
+												"running",
+												"building",
+												"pending",
+												"starting",
+												"probing",
+												"unhealthy",
+											].includes(s.status),
+										).length
+									}
+								</h2>
+							</div>
+						</div>
+						<SessionList
+							sessions={safeSessions}
+							onOpen={handleOpen}
+							onStop={handleStop}
+							onDelete={handleDelete}
+							onRecreateTunnel={handleRecreateTunnel}
+							actionBusyId={actionBusy}
+							tunnelHealth={tunnelHealth}
+							showGrouping={false}
+							emptyMessage="No active sessions right now."
+						/>
+					</section>
 
-      {status === "loading" && <LoadingState message="Loading overview..." />}
+					<section className="card stack home-section">
+						<div className="page-header">
+							<div>
+								<p className="eyebrow">Workspaces</p>
+								<h2>Quick access</h2>
+							</div>
+							<button
+								className="secondary-button"
+								type="button"
+								onClick={() => navigate("/workspaces")}
+							>
+								View all
+							</button>
+						</div>
+						<p className="muted">
+							Use the floating button to start a tool in any workspace.
+						</p>
+					</section>
 
-      {status === "error" && <ErrorState message="Unable to load your workspace overview." onRetry={() => void loadHome()} />}
-
-      {status === "ready" && summary && (
-        <>
-          <div className="home-summary-grid">
-            {summaryCards.map((card) => (
-              <article className="card home-summary-card" key={card.label}>
-                <p className="card-label">{card.label}</p>
-                <p className="card-value">
-                  {card.key === "openSessions"
-                    ? activeSessions.length
-                    : card.key === "projects"
-                      ? summary.projects
-                      : summary.repositories}
-                </p>
-              </article>
-            ))}
-          </div>
-
-          <section className="card stack home-section">
-            <div className="page-header">
-              <div>
-                <p className="eyebrow">Open sessions</p>
-                <h2>{safeSessions.filter((s) => ["running", "building", "pending", "starting", "probing", "unhealthy"].includes(s.status)).length}</h2>
-              </div>
-            </div>
-            <SessionList
-              sessions={safeSessions}
-              onOpen={handleOpen}
-              onStop={handleStop}
-              onDelete={handleDelete}
-              onRecreateTunnel={handleRecreateTunnel}
-              actionBusyId={actionBusy}
-              tunnelHealth={tunnelHealth}
-              showGrouping={false}
-              emptyMessage="No active sessions right now."
-            />
-          </section>
-
-          <section className="card stack home-section">
-            <div className="page-header">
-              <div>
-                <p className="eyebrow">Available projects</p>
-                <h2>{projects.length}</h2>
-              </div>
-              <button className="secondary-button" type="button" onClick={() => navigate("/projects")}>View all</button>
-            </div>
-            {projects.length === 0 ? (
-              <EmptyState message="No projects yet." />
-            ) : (
-              <div className="home-project-grid">
-                {projects.map((project) => (
-                  <article className="card project-card home-project-card" key={project.id}>
-                    <div className="stack-sm">
-                      <h3>{project.name}</h3>
-                      {project.description && <p className="muted">{project.description}</p>}
-                    </div>
-                    <button className="ghost-button small" type="button" onClick={() => navigate(`/projects/${project.id}`)}>
-                      Open Workspace
-                    </button>
-                  </article>
-                ))}
-              </div>
-            )}
-          </section>
-
-          <section className="card stack home-section">
-            <div className="page-header">
-              <div>
-                <p className="eyebrow">Quick create</p>
-                <h2>Start a session</h2>
-              </div>
-            </div>
-            <CreateSessionForm
-              projects={projects}
-              repositories={repositories}
-              toolTypes={toolTypes}
-              onProjectChange={(projectId) => setSelectedProject(projectId)}
-              onSuccess={handleCreateSuccess}
-            />
-          </section>
-
-          {safeSessions.filter((s) => ["stopped", "error"].includes(s.status)).length > 0 && (
-            <section className="card stack home-section">
-              <div className="page-header">
-                <div>
-                  <p className="eyebrow">Recent sessions</p>
-                  <h2>{safeSessions.filter((s) => ["stopped", "error"].includes(s.status)).length}</h2>
-                </div>
-              </div>
-              <SessionList
-                sessions={safeSessions}
-                onOpen={handleOpen}
-                onStart={handleStart}
-                onDelete={handleDelete}
-                actionBusyId={actionBusy}
-                showGrouping={false}
-                maxRecent={5}
-                emptyMessage="No recent sessions."
-              />
-            </section>
-          )}
-        </>
-      )}
-    </section>
-  );
+					{safeSessions.filter((s) => ["stopped", "error"].includes(s.status))
+						.length > 0 && (
+						<section className="card stack home-section">
+							<div className="page-header">
+								<div>
+									<p className="eyebrow">Recent sessions</p>
+									<h2>
+										{
+											safeSessions.filter((s) =>
+												["stopped", "error"].includes(s.status),
+											).length
+										}
+									</h2>
+								</div>
+							</div>
+							<SessionList
+								sessions={safeSessions}
+								onOpen={handleOpen}
+								onStart={handleStart}
+								onDelete={handleDelete}
+								actionBusyId={actionBusy}
+								showGrouping={false}
+								maxRecent={5}
+								emptyMessage="No recent sessions."
+							/>
+						</section>
+					)}
+				</>
+			)}
+		</section>
+	);
 };
 
 export { HomePage as DashboardPage };

apps/web/src/pages/projects.tsx

@@ -10,12 +10,12 @@ import {
 	type ProjectCreateInput,
 	type ProjectUpdateInput,
 } from "../api/projects";
+import { deleteWorkspace, syncWorkspace } from "../api/workspaces";
 import {
-	createWorkspace,
-	deleteWorkspace,
-	syncWorkspace,
-} from "../api/workspaces";
-import { EmptyState, ErrorState, LoadingState } from "../components/data-states";
+	EmptyState,
+	ErrorState,
+	LoadingState,
+} from "../components/data-states";
 import { Icon } from "../components/icon";
 import { WorkspaceCreateForm } from "../components/workspace-create-form";
 import { useAsyncData } from "../hooks/use-async-data";
@@ -24,10 +24,11 @@ import type { ProjectWithRepos, WorkspaceSummary } from "../types";
 type DialogMode = "none" | "create" | "edit";
 
 export const ProjectsPage = () => {
-	const { data: projects, status, reload } = useAsyncData<ProjectWithRepos[]>(
-		listProjects,
-		[],
-	);
+	const {
+		data: projects,
+		status,
+		reload,
+	} = useAsyncData<ProjectWithRepos[]>(listProjects, []);
 	const [dialogMode, setDialogMode] = useState<DialogMode>("none");
 	const [editingProject, setEditingProject] = useState<ProjectWithRepos | null>(
 		null,
@@ -107,23 +108,6 @@ export const ProjectsPage = () => {
 		}
 	};
 
-	const handleCreateWorkspace = async (
-		projectId: string,
-		repoId: string,
-		data: { name: string; branch: string },
-	) => {
-		setWorkspaceLoading(repoId);
-		try {
-			await createWorkspace(projectId, repoId, data);
-			setCreatingWorkspace(null);
-			reload();
-		} catch (err) {
-			alert(err instanceof Error ? err.message : "Failed to create workspace");
-		} finally {
-			setWorkspaceLoading(null);
-		}
-	};
-
 	const handleSyncWorkspace = async (
 		projectId: string,
 		repoId: string,
@@ -140,15 +124,11 @@ export const ProjectsPage = () => {
 		}
 	};
 
-	const handleDeleteWorkspace = async (
-		projectId: string,
-		repoId: string,
-		workspace: WorkspaceSummary,
-	) => {
+	const handleDeleteWorkspace = async (workspace: WorkspaceSummary) => {
 		if (!confirm(`Delete workspace "${workspace.name}"?`)) return;
 		setWorkspaceLoading(workspace.id);
 		try {
-			await deleteWorkspace(projectId, repoId, workspace.id);
+			await deleteWorkspace(workspace.id);
 			reload();
 		} catch (err) {
 			alert(err instanceof Error ? err.message : "Failed to delete workspace");
@@ -203,11 +183,7 @@ export const ProjectsPage = () => {
 								if (action === "sync") {
 									void handleSyncWorkspace(project.id, repoId, workspace);
 								} else if (action === "delete") {
-									void handleDeleteWorkspace(
-										project.id,
-										repoId,
-										workspace,
-									);
+									void handleDeleteWorkspace(workspace);
 								}
 							}}
 							workspaceLoading={workspaceLoading}
@@ -217,9 +193,10 @@ export const ProjectsPage = () => {
 									: null
 							}
 							onCancelCreate={() => setCreatingWorkspace(null)}
-							onSubmitCreate={async (repoId, data) =>
-								await handleCreateWorkspace(project.id, repoId, data)
-							}
+							onCreated={() => {
+								setCreatingWorkspace(null);
+								reload();
+							}}
 						/>
 					))}
 				</div>
@@ -298,7 +275,7 @@ function ProjectCard({
 	workspaceLoading,
 	showCreateForm,
 	onCancelCreate,
-	onSubmitCreate,
+	onCreated,
 }: {
 	project: ProjectWithRepos;
 	expanded: boolean;
@@ -317,7 +294,7 @@ function ProjectCard({
 	workspaceLoading: string | null;
 	onCancelCreate: () => void;
 	showCreateForm: string | null;
-	onSubmitCreate: (repoId: string, data: { name: string; branch: string }) => Promise<void>;
+	onCreated: () => void;
 }) {
 	return (
 		<article className="card project-card">
@@ -328,10 +305,7 @@ function ProjectCard({
 					type="button"
 					aria-expanded={expanded}
 				>
-					<Icon
-						name={expanded ? "chevron-down" : "chevron-right"}
-						size="sm"
-					/>
+					<Icon name={expanded ? "chevron-down" : "chevron-right"} size="sm" />
 					<h3>{project.name}</h3>
 					{project.repositories.length > 0 && (
 						<span className="repo-count">
@@ -398,11 +372,9 @@ function ProjectCard({
 									</div>
 									{showCreateForm === repo.id && (
 										<WorkspaceCreateForm
-											projectId={project.id}
-											repoId={repo.id}
-											onSubmit={(data) =>
-												onSubmitCreate(repo.id, data)
-											}
+											defaultProjectId={project.id}
+											defaultRepoId={repo.id}
+											onSubmit={onCreated}
 											onCancel={onCancelCreate}
 										/>
 									)}
@@ -428,15 +400,9 @@ function ProjectCard({
 													<div className="ws-actions">
 														<button
 															type="button"
-															disabled={
-																workspaceLoading === ws.id
-															}
+															disabled={workspaceLoading === ws.id}
 															onClick={() =>
-																onWorkspaceAction(
-																	repo.id,
-																	ws,
-																	"sync",
-																)
+																onWorkspaceAction(repo.id, ws, "sync")
 															}
 														>
 															<Icon name="refresh" size="sm" />
@@ -444,15 +410,9 @@ function ProjectCard({
 														<button
 															type="button"
 															className="danger-text"
-															disabled={
-																workspaceLoading === ws.id
-															}
+															disabled={workspaceLoading === ws.id}
 															onClick={() =>
-																onWorkspaceAction(
-																	repo.id,
-																	ws,
-																	"delete",
-																)
+																onWorkspaceAction(repo.id, ws, "delete")
 															}
 														>
 															<Icon name="delete" size="sm" />

apps/web/src/pages/sessions.tsx

@@ -1,17 +1,12 @@
-import { useCallback, useEffect, useMemo, useState } from "react";
+import { useCallback, useEffect, useState } from "react";
 
-import { listProjects } from "../api/projects";
-import type { ProjectWithRepos } from "../types";
-import { listRepositories, type GitRepository } from "../api/git_repositories";
 import {
-  getUserSessions,
-  type Session,
-  checkInstanceHealth,
+	getUserSessions,
+	type Session,
+	checkInstanceHealth,
 } from "../api/sessions";
-import { listToolTypes, type ToolType } from "../api/tool_types";
-import { getUserConfig, updateUserConfig } from "../api/settings";
+import { getUserConfig } from "../api/settings";
 import { ErrorState, LoadingState } from "../components/data-states";
-import { CreateSessionForm } from "../components/create-session-form";
 import { SessionList } from "../components/session-list";
 import { SessionCard } from "../components/session-card";
 import { useInstanceActions } from "../hooks/use-instance-actions";
@@ -20,235 +15,187 @@ import type { InstanceHealth } from "../api/sessions";
 type SessionsStatus = "loading" | "ready" | "error";
 
 export const SessionsPage = () => {
-  const [status, setStatus] = useState<SessionsStatus>("loading");
-  const [sessions, setSessions] = useState<Session[]>([]);
-  const [lastSessionId, setLastSessionId] = useState<string | null>(null);
+	const [status, setStatus] = useState<SessionsStatus>("loading");
+	const [sessions, setSessions] = useState<Session[]>([]);
+	const [lastSessionId, setLastSessionId] = useState<string | null>(null);
 
-  const [projects, setProjects] = useState<ProjectWithRepos[]>([]);
-  const [repositories, setRepositories] = useState<GitRepository[]>([]);
-  const [toolTypes, setToolTypes] = useState<ToolType[]>([]);
-  const [selectedProject, setSelectedProject] = useState<string>("");
+	const [tunnelHealth, setTunnelHealth] = useState<
+		Record<string, InstanceHealth>
+	>({});
 
-  const [tunnelHealth, setTunnelHealth] = useState<Record<string, InstanceHealth>>({});
+	const loadSessions = useCallback(async () => {
+		setStatus("loading");
+		try {
+			const [sessionsData, config] = await Promise.all([
+				getUserSessions(),
+				getUserConfig(),
+			]);
+			setSessions(sessionsData);
+			setLastSessionId(config.last_session_id ?? null);
+			setStatus("ready");
+		} catch {
+			setStatus("error");
+		}
+	}, []);
 
-  const loadSessions = useCallback(async () => {
-    setStatus("loading");
-    try {
-      const [sessionsData, config] = await Promise.all([
-        getUserSessions(),
-        getUserConfig(),
-      ]);
-      setSessions(sessionsData);
-      setLastSessionId(config.last_session_id ?? null);
-      setStatus("ready");
-    } catch {
-      setStatus("error");
-    }
-  }, []);
+	useEffect(() => {
+		void loadSessions();
+	}, [loadSessions]);
 
-  useEffect(() => {
-    void loadSessions();
-  }, [loadSessions]);
+	const {
+		loadingSessionId,
+		dirtyDeleteSession,
+		dirtyDeleteFiles,
+		handleOpen,
+		handleStart,
+		handleStop,
+		handleDelete,
+		handleForceDelete,
+		handleRecreateTunnel,
+		clearDirtyDelete,
+	} = useInstanceActions({ onRefresh: loadSessions });
 
-  useEffect(() => {
-    const loadProjects = async () => {
-      try {
-        const data = await listProjects();
-        setProjects(data);
-      } catch {
-        // ignore
-      }
-    };
-    void loadProjects();
-  }, []);
+	// Poll health every 30 seconds for active web-enabled instances
+	useEffect(() => {
+		const checkHealth = async () => {
+			const activeSessions = sessions.filter(
+				(s) =>
+					["running", "starting", "unhealthy", "probing"].includes(s.status) &&
+					s.tool_type_interfaces?.includes("web"),
+			);
+			for (const session of activeSessions) {
+				try {
+					const health = await checkInstanceHealth(
+						session.project_id,
+						session.repository_id,
+						session.id,
+					);
+					setTunnelHealth((prev) => ({
+						...prev,
+						[session.id]: health,
+					}));
+				} catch {
+					setTunnelHealth((prev) => ({
+						...prev,
+						[session.id]: {
+							healthy: false,
+							container_status: "unknown",
+							container_health: null,
+							tunnel_status: "unreachable",
+							tunnel_status_code: null,
+							probe_status: "unknown",
+							last_probe_output: null,
+							error: "check failed",
+						} as InstanceHealth,
+					}));
+				}
+			}
+		};
 
-  useEffect(() => {
-    const loadToolTypes = async () => {
-      try {
-        const data = await listToolTypes();
-        setToolTypes(data);
-      } catch {
-        // ignore
-      }
-    };
-    void loadToolTypes();
-  }, []);
+		void checkHealth();
+		const interval = setInterval(() => void checkHealth(), 30000);
+		return () => clearInterval(interval);
+	}, [sessions]);
 
-  const {
-    loadingSessionId,
-    dirtyDeleteSession,
-    dirtyDeleteFiles,
-    handleOpen,
-    handleStart,
-    handleStop,
-    handleDelete,
-    handleForceDelete,
-    handleRecreateTunnel,
-    clearDirtyDelete,
-  } = useInstanceActions({ onRefresh: loadSessions });
+	const lastSession = sessions.find((s) => s.id === lastSessionId) ?? null;
 
-  // Poll health every 30 seconds for active web-enabled instances
-  useEffect(() => {
-    const checkHealth = async () => {
-      const activeSessions = sessions.filter(
-        (s) => ["running", "starting", "unhealthy", "probing"].includes(s.status)
-          && s.tool_type_interfaces?.includes("web")
-      );
-      for (const session of activeSessions) {
-        try {
-          const health = await checkInstanceHealth(
-            session.project_id,
-            session.repository_id,
-            session.id
-          );
-          setTunnelHealth((prev) => ({
-            ...prev,
-            [session.id]: health,
-          }));
-        } catch {
-          setTunnelHealth((prev) => ({
-            ...prev,
-            [session.id]: {
-              healthy: false,
-              container_status: "unknown",
-              container_health: null,
-              tunnel_status: "unreachable",
-              tunnel_status_code: null,
-              probe_status: "unknown",
-              last_probe_output: null,
-              error: "check failed",
-            } as InstanceHealth,
-          }));
-        }
-      }
-    };
+	return (
+		<section className="stack sessions-page">
+			<div className="page-header">
+				<h1>Sessions</h1>
+			</div>
 
-    // Check immediately and then every 30 seconds
-    void checkHealth();
-    const interval = setInterval(() => void checkHealth(), 30000);
-    return () => clearInterval(interval);
-  }, [sessions]);
+			{status === "loading" && <LoadingState message="Loading sessions..." />}
 
-  useEffect(() => {
-    if (!selectedProject) {
-      setRepositories([]);
-      return;
-    }
-    const loadRepos = async () => {
-      try {
-        const data = await listRepositories(selectedProject);
-        setRepositories(data);
-      } catch {
-        setRepositories([]);
-      }
-    };
-    void loadRepos();
-  }, [selectedProject]);
+			{status === "error" && (
+				<ErrorState
+					message="Failed to load sessions"
+					onRetry={() => void loadSessions()}
+				/>
+			)}
 
-  const lastSession = useMemo(
-    () => sessions.find((s) => s.id === lastSessionId) ?? null,
-    [sessions, lastSessionId]
-  );
+			{status === "ready" && (
+				<>
+					{/* Last Session */}
+					{lastSession && (
+						<div className="last-session-section">
+							<h2>Last Session</h2>
+							<SessionCard
+								session={lastSession}
+								onOpen={handleOpen}
+								onDelete={handleDelete}
+								isBusy={loadingSessionId === lastSession.id}
+								tunnelHealth={tunnelHealth[lastSession.id] || null}
+							/>
+						</div>
+					)}
 
-  const handleCreateSuccess = async (instance: { id: string }) => {
-    await updateUserConfig({ last_session_id: instance.id });
-    setSelectedProject("");
-    await loadSessions();
-  };
+					{/* Session List */}
+					<div className="sessions-list-wrapper">
+						<SessionList
+							sessions={sessions}
+							onOpen={handleOpen}
+							onStart={handleStart}
+							onStop={handleStop}
+							onDelete={handleDelete}
+							onRecreateTunnel={handleRecreateTunnel}
+							actionBusyId={loadingSessionId}
+							tunnelHealth={tunnelHealth}
+						/>
+					</div>
 
-  return (
-    <section className="stack sessions-page">
-      <div className="page-header">
-        <h1>Sessions</h1>
-      </div>
+					{/* Create Session — use floating button */}
+					<div className="create-session-section">
+						<h2>Start New Tool</h2>
+						<p className="muted">
+							Use the floating button (bottom-right) to start a tool in any
+							workspace.
+						</p>
+					</div>
 
-      {status === "loading" && <LoadingState message="Loading sessions..." />}
-
-      {status === "error" && <ErrorState message="Failed to load sessions" onRetry={() => void loadSessions()} />}
-
-      {status === "ready" && (
-        <>
-          {/* Last Session */}
-          {lastSession && (
-            <div className="last-session-section">
-              <h2>Last Session</h2>
-              <SessionCard
-                session={lastSession}
-                onOpen={handleOpen}
-                onDelete={handleDelete}
-                isBusy={loadingSessionId === lastSession.id}
-                tunnelHealth={tunnelHealth[lastSession.id] || null}
-              />
-            </div>
-          )}
-
-          {/* Session List */}
-          <div className="sessions-list-wrapper">
-            <SessionList
-              sessions={sessions}
-              onOpen={handleOpen}
-              onStart={handleStart}
-              onStop={handleStop}
-              onDelete={handleDelete}
-              onRecreateTunnel={handleRecreateTunnel}
-              actionBusyId={loadingSessionId}
-              tunnelHealth={tunnelHealth}
-            />
-          </div>
-
-          {/* Create Session */}
-          <div className="create-session-section">
-            <h2>Create New Session</h2>
-            <CreateSessionForm
-              projects={projects}
-              repositories={repositories}
-              toolTypes={toolTypes}
-              onProjectChange={(projectId) => {
-                setSelectedProject(projectId);
-              }}
-              onSuccess={handleCreateSuccess}
-            />
-          </div>
-
-          {/* Dirty Delete Confirmation Modal */}
-          {dirtyDeleteSession && (
-            <div className="modal-overlay" onClick={clearDirtyDelete}>
-              <div className="modal-content" onClick={(e) => e.stopPropagation()}>
-                <h3>Uncommitted Changes</h3>
-                <p>
-                  The repository <strong>{dirtyDeleteSession.repository_name}</strong> has
-                  uncommitted changes. Deleting this session will permanently lose these
-                  changes.
-                </p>
-                <div className="changed-files-list">
-                  <h4>Changed files:</h4>
-                  <ul>
-                    {dirtyDeleteFiles.map((file, idx) => (
-                      <li key={idx}>{file}</li>
-                    ))}
-                  </ul>
-                </div>
-                <div className="modal-actions">
-                  <button
-                    className="secondary-button"
-                    onClick={clearDirtyDelete}
-                    type="button"
-                  >
-                    Cancel
-                  </button>
-                  <button
-                    className="danger-button"
-                    onClick={() => void handleForceDelete(dirtyDeleteSession)}
-                    type="button"
-                  >
-                    Force Delete
-                  </button>
-                </div>
-              </div>
-            </div>
-          )}
-        </>
-      )}
-    </section>
-  );
+					{/* Dirty Delete Confirmation Modal */}
+					{dirtyDeleteSession && (
+						<div className="modal-overlay" onClick={clearDirtyDelete}>
+							<div
+								className="modal-content"
+								onClick={(e) => e.stopPropagation()}
+							>
+								<h3>Uncommitted Changes</h3>
+								<p>
+									The repository{" "}
+									<strong>{dirtyDeleteSession.repository_name}</strong> has
+									uncommitted changes. Deleting this session will permanently
+									lose these changes.
+								</p>
+								<div className="changed-files-list">
+									<h4>Changed files:</h4>
+									<ul>
+										{dirtyDeleteFiles.map((file, idx) => (
+											<li key={idx}>{file}</li>
+										))}
+									</ul>
+								</div>
+								<div className="modal-actions">
+									<button
+										className="secondary-button"
+										onClick={clearDirtyDelete}
+										type="button"
+									>
+										Cancel
+									</button>
+									<button
+										className="danger-button"
+										onClick={() => void handleForceDelete(dirtyDeleteSession)}
+										type="button"
+									>
+										Force Delete
+									</button>
+								</div>
+							</div>
+						</div>
+					)}
+				</>
+			)}
+		</section>
+	);
 };

apps/web/src/pages/workspace-detail.tsx

@@ -8,7 +8,9 @@ import { useWorkspaceFiles } from "../hooks/use-workspace-files";
 import { useWorkspaceGit } from "../hooks/use-workspace-git";
 import { useWorkspaceInstances } from "../hooks/use-workspace-instances";
 import { useMobileViewport } from "../hooks/use-mobile-viewport";
+import { ToolStarter } from "../components/tool-starter";
 import type { FileEntry } from "../api/workspace-files";
+import type { Workspace } from "../types/workspace";
 
 type Tab = "files" | "git" | "tools" | "settings";
 
@@ -40,7 +42,7 @@ export function WorkspaceDetailPage() {
 			<div className="workspace-content">
 				{activeTab === "files" && <FilesTab workspaceId={workspace.id} />}
 				{activeTab === "git" && <GitTab workspaceId={workspace.id} />}
-				{activeTab === "tools" && <ToolsTab workspaceId={workspace.id} />}
+				{activeTab === "tools" && <ToolsTab workspace={workspace} />}
 				{activeTab === "settings" && <SettingsTab workspace={workspace} />}
 			</div>
 			{isMobile && <MobileTabBar active={activeTab} onChange={setActiveTab} />}
@@ -100,7 +102,10 @@ function TabBar({
 					role="tab"
 					aria-selected={active === tab.id}
 				>
-					<Icon name={tab.icon as "folder" | "branch" | "terminal" | "settings"} size="sm" />
+					<Icon
+						name={tab.icon as "folder" | "branch" | "terminal" | "settings"}
+						size="sm"
+					/>
 					{tab.label}
 				</button>
 			))}
@@ -132,7 +137,9 @@ function MobileTabBar({
 					role="tab"
 					aria-selected={active === tab.id}
 				>
-					<Icon name={tab.icon as "folder" | "branch" | "terminal" | "settings"} />
+					<Icon
+						name={tab.icon as "folder" | "branch" | "terminal" | "settings"}
+					/>
 					<span>{tab.label}</span>
 				</button>
 			))}
@@ -304,8 +311,8 @@ function GitTab({ workspaceId }: { workspaceId: string }) {
 
 /* ─── Tools Tab ─── */
 
-function ToolsTab({ workspaceId }: { workspaceId: string }) {
-	const { instances, loading, create } = useWorkspaceInstances(workspaceId);
+function ToolsTab({ workspace }: { workspace: Workspace }) {
+	const { instances, loading, refresh } = useWorkspaceInstances(workspace.id);
 	const [showModal, setShowModal] = useState(false);
 
 	return (
@@ -354,13 +361,19 @@ function ToolsTab({ workspaceId }: { workspaceId: string }) {
 				</>
 			)}
 			{showModal && (
-				<StartToolModal
-					onClose={() => setShowModal(false)}
-					onStart={async (toolTypeId: string) => {
-						await create(toolTypeId);
-						setShowModal(false);
-					}}
-				/>
+				<div className="modal-overlay" onClick={() => setShowModal(false)}>
+					<div className="modal-content" onClick={(e) => e.stopPropagation()}>
+						<h3>Start Tool</h3>
+						<ToolStarter
+							workspace={workspace}
+							onStarted={() => {
+								setShowModal(false);
+								void refresh();
+							}}
+							onCancel={() => setShowModal(false)}
+						/>
+					</div>
+				</div>
 			)}
 		</div>
 	);
@@ -368,18 +381,7 @@ function ToolsTab({ workspaceId }: { workspaceId: string }) {
 
 /* ─── Settings Tab ─── */
 
-function SettingsTab({
-	workspace,
-}: {
-	workspace: {
-		id: string;
-		name: string;
-		branch: string;
-		path: string;
-		status: string;
-		created_at: string;
-	};
-}) {
+function SettingsTab({ workspace }: { workspace: Workspace }) {
 	return (
 		<div className="settings-tab">
 			<div className="settings-section">
@@ -410,57 +412,3 @@ function SettingsTab({
 		</div>
 	);
 }
-
-/* ─── Start Tool Modal ─── */
-
-function StartToolModal({
-	onClose,
-	onStart,
-}: {
-	onClose: () => void;
-	onStart: (toolTypeId: string) => Promise<void>;
-}) {
-	const [toolTypeId, setToolTypeId] = useState("");
-	const [submitting, setSubmitting] = useState(false);
-
-	const handleSubmit = async (e: React.FormEvent) => {
-		e.preventDefault();
-		if (!toolTypeId) return;
-		setSubmitting(true);
-		try {
-			await onStart(toolTypeId);
-		} finally {
-			setSubmitting(false);
-		}
-	};
-
-	return (
-		<div className="modal-overlay" onClick={onClose}>
-			<div className="modal-content" onClick={(e) => e.stopPropagation()}>
-				<h3>Start Tool</h3>
-				<form onSubmit={handleSubmit}>
-					<div className="form-group">
-						<label>Tool Type</label>
-						<select
-							value={toolTypeId}
-							onChange={(e) => setToolTypeId(e.target.value)}
-						>
-							<option value="">Select...</option>
-							<option value="code-server">Code Server</option>
-							<option value="jupyter-notebook">Jupyter Notebook</option>
-							<option value="terminal">Terminal</option>
-						</select>
-					</div>
-					<div className="form-actions">
-						<button type="button" onClick={onClose} disabled={submitting}>
-							Cancel
-						</button>
-						<button type="submit" disabled={!toolTypeId || submitting}>
-							{submitting ? "Starting..." : "Start"}
-						</button>
-					</div>
-				</form>
-			</div>
-		</div>
-	);
-}

apps/web/src/pages/workspaces.tsx

@@ -6,36 +6,17 @@ import { useWorkspaces } from "../hooks/use-workspaces";
 import { useWorkspaceActions } from "../hooks/use-workspace-actions";
 import { WorkspaceCard } from "../components/workspace-card";
 import { WorkspaceCreateForm } from "../components/workspace-create-form";
-import { StartToolModal } from "../components/start-tool-modal";
-import { createInstance, startInstance } from "../api/sessions";
+import { ToolStarter } from "../components/tool-starter";
 import type { Workspace } from "../types/workspace";
 
 export function WorkspacesPage() {
 	const [showCreate, setShowCreate] = useState(false);
 	const [startWorkspace, setStartWorkspace] = useState<Workspace | null>(null);
-	const [createTarget, setCreateTarget] = useState<{
-		projectId: string;
-		repoId: string;
-	} | null>(null);
 
 	const { workspaces, loading, error, refresh } = useWorkspaces();
 	const actions = useWorkspaceActions();
-
-	const handleCreate = async (data: { name: string; branch: string }) => {
-		if (!createTarget) return;
-		await actions.create(createTarget.projectId, createTarget.repoId, data);
-		setShowCreate(false);
-		setCreateTarget(null);
-		await refresh();
-	};
-
 	const handleDelete = async (workspace: Workspace) => {
-		await actions.delete(
-			workspace.project_id,
-			workspace.repo_id,
-			workspace,
-			refresh,
-		);
+		await actions.delete(workspace, refresh);
 	};
 
 	const handleSync = async (workspace: Workspace) => {
@@ -47,37 +28,6 @@ export function WorkspacesPage() {
 		);
 	};
 
-	const handleStartTool = async (
-		toolTypeId: string,
-		configProfileId?: string,
-	) => {
-		if (!startWorkspace) return;
-		try {
-			const instance = await createInstance(
-				startWorkspace.project_id,
-				startWorkspace.repo_id,
-				toolTypeId,
-				`${startWorkspace.name} - ${toolTypeId}`,
-				undefined,
-				undefined,
-				undefined,
-				configProfileId,
-				[],
-				startWorkspace.id,
-			);
-			await startInstance(
-				startWorkspace.project_id,
-				startWorkspace.repo_id,
-				instance.id,
-				configProfileId,
-			);
-			setStartWorkspace(null);
-			await refresh();
-		} catch (err) {
-			alert(err instanceof Error ? err.message : "Failed to start tool");
-		}
-	};
-
 	return (
 		<div className="page workspaces-page">
 			<header className="page-header">
@@ -92,18 +42,7 @@ export function WorkspacesPage() {
 					</button>
 					<button
 						className="btn btn-primary"
-						onClick={() => {
-							if (workspaces.length > 0) {
-								const first = workspaces[0];
-								setCreateTarget({
-									projectId: first.project_id,
-									repoId: first.repo_id,
-								});
-								setShowCreate(true);
-							} else {
-								alert("Navigate to a project to create your first workspace.");
-							}
-						}}
+						onClick={() => setShowCreate(true)}
 					>
 						<Icon name="add" size="sm" /> New Workspace
 					</button>
@@ -112,15 +51,13 @@ export function WorkspacesPage() {
 
 			{error && <div className="alert alert-error">{error}</div>}
 
-			{showCreate && createTarget && (
+			{showCreate && (
 				<WorkspaceCreateForm
-					projectId={createTarget.projectId}
-					repoId={createTarget.repoId}
-					onSubmit={handleCreate}
-					onCancel={() => {
+					onSubmit={async () => {
 						setShowCreate(false);
-						setCreateTarget(null);
+						await refresh();
 					}}
+					onCancel={() => setShowCreate(false)}
 				/>
 			)}
 
@@ -129,7 +66,12 @@ export function WorkspacesPage() {
 			) : workspaces.length === 0 ? (
 				<div className="empty-state">
 					<p>No workspaces yet.</p>
-					<p>Navigate to a project to create your first workspace.</p>
+					<button
+						className="btn btn-primary"
+						onClick={() => setShowCreate(true)}
+					>
+						<Icon name="add" size="sm" /> Create your first workspace
+					</button>
 				</div>
 			) : (
 				<div className="workspaces-grid">
@@ -147,11 +89,19 @@ export function WorkspacesPage() {
 			)}
 
 			{startWorkspace && (
-				<StartToolModal
-					workspace={startWorkspace}
-					onClose={() => setStartWorkspace(null)}
-					onStart={handleStartTool}
-				/>
+				<div className="modal-overlay" onClick={() => setStartWorkspace(null)}>
+					<div className="modal-content" onClick={(e) => e.stopPropagation()}>
+						<h3>Start Tool</h3>
+						<ToolStarter
+							workspace={startWorkspace}
+							onStarted={() => {
+								setStartWorkspace(null);
+								void refresh();
+							}}
+							onCancel={() => setStartWorkspace(null)}
+						/>
+					</div>
+				</div>
 			)}
 		</div>
 	);

apps/web/src/router.tsx

@@ -58,7 +58,10 @@ export const AppRouter = () => {
 				</Route>
 				<Route path="sessions" element={<SessionsPage />} />
 				<Route path="workspaces" element={<WorkspacesPage />} />
-				<Route path="workspaces/:workspaceId" element={<WorkspaceDetailPage />} />
+				<Route
+					path="workspaces/:workspaceId"
+					element={<WorkspaceDetailPage />}
+				/>
 				<Route path="tool-workshop" element={<ToolWorkshopPage />} />
 				<Route
 					path="instances/:instanceId/terminal"

apps/web/src/styles.css

@@ -5454,3 +5454,220 @@ a:active,
 .workspace-chip .ws-actions button.danger-text:hover {
 	color: var(--danger);
 }
+
+/* ─── Workspace Create Inline ─── */
+
+.workspace-create-inline {
+	padding: var(--space-5);
+	margin-bottom: var(--space-5);
+}
+
+.workspace-create-inline h3 {
+	margin: 0 0 var(--space-4) 0;
+	display: flex;
+	align-items: center;
+	gap: var(--space-2);
+}
+
+.workspace-create-form-grid {
+	display: grid;
+	grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
+	gap: var(--space-4);
+	align-items: end;
+}
+
+.workspace-create-form-grid .form-group {
+	display: flex;
+	flex-direction: column;
+	gap: var(--space-1);
+}
+
+.workspace-create-form-grid label {
+	font-size: var(--font-size-sm);
+	color: var(--muted);
+}
+
+.workspace-create-form-grid input,
+.workspace-create-form-grid select {
+	padding: var(--space-2) var(--space-3);
+	border: 1px solid var(--border);
+	border-radius: 8px;
+	font: inherit;
+	background: var(--panel);
+	color: var(--ink);
+}
+
+.workspace-create-form-grid .form-actions {
+	display: flex;
+	gap: var(--space-3);
+	justify-content: flex-end;
+	margin-top: var(--space-2);
+}
+
+/* ─── Tool Starter ─── */
+.tool-starter {
+	display: flex;
+	flex-direction: column;
+	gap: var(--space-4);
+}
+
+.tool-starter-context {
+	background: var(--bg);
+	border: 1px solid var(--border);
+	border-radius: 10px;
+	padding: var(--space-3);
+	display: flex;
+	flex-direction: column;
+	gap: var(--space-2);
+}
+
+.context-row {
+	display: flex;
+	align-items: center;
+	gap: var(--space-3);
+	flex-wrap: wrap;
+}
+
+.context-label {
+	font-size: var(--font-size-sm);
+	color: var(--muted);
+	min-width: 80px;
+}
+
+.context-value {
+	font-weight: 500;
+}
+
+.ssh-key-status {
+	padding: var(--space-2);
+	background: var(--bg);
+	border-radius: 8px;
+}
+
+.ssh-key-status .success-text,
+.ssh-key-status .warning-text {
+	display: flex;
+	align-items: center;
+	gap: var(--space-2);
+	font-size: var(--font-size-sm);
+}
+
+/* ─── Instance Chips ─── */
+.instance-chips {
+	display: flex;
+	flex-wrap: wrap;
+	gap: var(--space-2);
+	margin-top: var(--space-2);
+}
+
+.instance-chip {
+	display: inline-flex;
+	align-items: center;
+	gap: var(--space-1);
+	padding: 0.2rem 0.5rem;
+	border-radius: 999px;
+	font-size: var(--font-size-sm);
+	background: var(--bg);
+	border: 1px solid var(--border);
+}
+
+.instance-chip.running {
+	background: var(--success-light);
+	border-color: var(--success);
+	color: var(--success);
+}
+
+.instance-chip.starting,
+.instance-chip.building,
+.instance-chip.probing {
+	background: var(--warning-light);
+	border-color: var(--warning);
+	color: var(--warning);
+}
+
+.instance-chip.error,
+.instance-chip.unhealthy,
+.instance-chip.stopped {
+	background: var(--danger-light);
+	border-color: var(--danger);
+	color: var(--danger);
+}
+
+.instance-chip a {
+	font-size: 0.75rem;
+	opacity: 0.7;
+}
+
+.instance-chip a:hover {
+	opacity: 1;
+}
+
+/* ─── Tool Starter Header ─── */
+.tool-starter-header {
+	display: flex;
+	justify-content: space-between;
+	align-items: center;
+	margin-bottom: var(--space-3);
+	padding-bottom: var(--space-2);
+	border-bottom: 1px solid var(--border);
+}
+
+.tool-starter-header h4 {
+	margin: 0;
+}
+
+/* ─── Floating Action Button ─── */
+.start-tool-fab {
+	position: fixed;
+	bottom: 2rem;
+	right: 2rem;
+	z-index: 100;
+	width: 3.5rem;
+	height: 3.5rem;
+	border-radius: 50%;
+	background: var(--primary);
+	color: var(--primary-fg);
+	border: none;
+	box-shadow: 0 4px 14px rgba(0, 0, 0, 0.25);
+	cursor: pointer;
+	display: flex;
+	align-items: center;
+	justify-content: center;
+	transition:
+		transform 0.15s ease,
+		box-shadow 0.15s ease;
+}
+
+.start-tool-fab:hover {
+	transform: scale(1.08);
+	box-shadow: 0 6px 20px rgba(0, 0, 0, 0.3);
+}
+
+.start-tool-fab:active {
+	transform: scale(0.95);
+}
+
+.start-tool-modal {
+	max-width: 480px;
+	width: 90vw;
+}
+
+.modal-header {
+	display: flex;
+	justify-content: space-between;
+	align-items: center;
+	margin-bottom: var(--space-3);
+}
+
+.modal-header h3 {
+	margin: 0;
+}
+
+@media (max-width: 640px) {
+	.start-tool-fab {
+		bottom: 5rem; /* above mobile tab bar */
+		right: 1rem;
+		width: 3rem;
+		height: 3rem;
+	}
+}

apps/web/src/types.ts

@@ -1,43 +1,43 @@
 export type SessionUser = {
-  id: string;
-  email: string;
-  name: string;
-  avatar_url: string | null;
+	id: string;
+	email: string;
+	name: string;
+	avatar_url: string | null;
 };
 
 export type SessionPayload = {
-  user: SessionUser;
+	user: SessionUser;
 };
 
 export type Project = {
-  id: string;
-  name: string;
-  description: string | null;
-  owner_id: string;
-  default_ssh_key_id: string | null;
+	id: string;
+	name: string;
+	description: string | null;
+	owner_id: string;
+	default_ssh_key_id: string | null;
 };
 
 export type WorkspaceSummary = {
-  id: string;
-  name: string;
-  branch: string;
-  status: string;
-  instance_count: number;
+	id: string;
+	name: string;
+	branch: string;
+	status: string;
+	instance_count: number;
 };
 
 export type RepositorySummary = {
-  id: string;
-  name: string;
-  remote_url: string;
-  workspaces: WorkspaceSummary[];
+	id: string;
+	name: string;
+	remote_url: string;
+	workspaces: WorkspaceSummary[];
 };
 
 export type ProjectWithRepos = {
-  id: string;
-  name: string;
-  description: string | null;
-  owner_id: string;
-  default_ssh_key_id: string | null;
-  repositories: RepositorySummary[];
-  created_at: string;
+	id: string;
+	name: string;
+	description: string | null;
+	owner_id: string;
+	default_ssh_key_id: string | null;
+	repositories: RepositorySummary[];
+	created_at: string;
 };

apps/web/src/types/workspace.ts

@@ -5,6 +5,7 @@ export interface Workspace {
 	name: string;
 	repo_id: string;
 	repo_name: string;
+	repo_ssh_key_id: string | null;
 	project_id: string;
 	project_name: string;
 	user_id: string;

docker-compose.traefik.yml

@@ -13,7 +13,11 @@ services:
     volumes:
       - postgres_data:/var/lib/postgresql/data
     healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-headquarter} -d ${POSTGRES_DB:-headquarter}"]
+      test:
+        [
+          "CMD-SHELL",
+          "pg_isready -U ${POSTGRES_USER:-headquarter} -d ${POSTGRES_DB:-headquarter}",
+        ]
       interval: 10s
       timeout: 5s
       retries: 5
@@ -93,6 +97,7 @@ services:
       AUTHENTIK_TOKEN_URL: ${AUTHENTIK_TOKEN_URL:-}
     volumes:
       - /data/repos:/data/repos
+      - /data/working-copies:/data/working-copies
       - /data/instances:/data/instances
       - avatar_uploads:/app/uploads
       - /var/run/docker.sock:/var/run/docker.sock

docker-compose.yml

@@ -62,6 +62,7 @@ services:
       INSTANCE_BASE_PATH: /data/instances
     volumes:
       - /data/repos:/data/repos
+      - /data/working-copies:/data/working-copies
       - /data/instances:/data/instances
     ports:
       - "8000:8000"

openspec/designs/terminal-responsiveness.md

@@ -0,0 +1,227 @@
+# Design: High-Performance Web Terminal
+
+## Component Diagram
+
+```
+┌─────────────────────────────────────────────────────────────┐
+│                        Browser                              │
+│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────┐ │
+│  │ Terminal    │  │ WebSocket   │  │ xterm.js            │ │
+│  │ Component   │──│ Client      │──│ + WebGL addon       │ │
+│  │             │  │ (binary)    │  │ + DOM fallback      │ │
+│  └─────────────┘  └─────────────┘  └─────────────────────┘ │
+│         │                                              │     │
+│         │ Flow control ack                             │     │
+│         └──────────────────────────────────────────────┘     │
+└─────────────────────────────────────────────────────────────┘
+                              │ WebSocket
+                              │ (binary frames)
+┌─────────────────────────────────────────────────────────────┐
+│                      API Container                          │
+│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────┐ │
+│  │ Terminal    │  │ WebSocket   │  │ TerminalSession     │ │
+│  │ Manager     │──│ Endpoint    │──│ (new)               │ │
+│  │ (lifecycle) │  │ (router)    │  │ - asyncio fd reader │ │
+│  └─────────────┘  └─────────────┘  │ - output batcher    │ │
+│                                     │ - flow control      │ │
+│                                     └─────────────────────┘ │
+│                                              │               │
+│                                     ┌────────┴────────┐      │
+│                                     │ docker exec -it │      │
+│                                     │ (subprocess)    │      │
+│                                     └────────┬────────┘      │
+│                                              │               │
+│                                     ┌────────┴────────┐      │
+│                                     │ Tool Container  │      │
+│                                     │ (bash shell)    │      │
+│                                     └─────────────────┘      │
+└─────────────────────────────────────────────────────────────┘
+```
+
+## File Changes
+
+### New Files
+- `apps/api/src/services/terminal_session_v2.py` — New TerminalSession implementation
+
+### Modified Files
+- `apps/api/src/services/terminal_session.py` — Delete (or keep as legacy, user said no legacy needed)
+- `apps/api/src/services/terminal_manager.py` — Update to use new TerminalSession
+- `apps/api/src/api/terminal.py` — Update WebSocket handler for binary frames + flow control
+- `apps/web/src/components/terminal.tsx` — Binary mode, WebGL, flow control ack
+- `apps/web/package.json` — Add `xterm-addon-webgl`
+
+## TerminalSession Implementation
+
+```python
+class TerminalSession:
+    """High-performance terminal session with asyncio-native I/O."""
+    
+    BUFFER_SIZE = 10 * 1024
+    IDLE_TIMEOUT = 30 * 60
+    BATCH_WINDOW_MS = 2
+    FLOW_CONTROL_THRESHOLD = 64 * 1024
+    FLOW_CONTROL_RESUME = 32 * 1024
+    
+    def __init__(self, session_id, instance_id, container_id, ...):
+        self._master_fd: int | None = None
+        self._process: asyncio.subprocess.Process | None = None
+        self._output_buffer: deque[bytes] = deque(maxlen=self.BUFFER_SIZE)
+        self._websockets: set[WebSocket] = set()
+        self._batch_buffer = bytearray()
+        self._batch_timer: asyncio.TimerHandle | None = None
+        self._unacknowledged_bytes = 0
+        self._paused = False
+        self._read_handler_set = False
+        
+    async def start(self):
+        self._master_fd, slave_fd = pty.openpty()
+        self._set_terminal_size(80, 24)
+        
+        self._process = await asyncio.create_subprocess_exec(
+            "docker", "exec", "-it", "-e", "TERM=xterm-256color",
+            self.container_id, "bash", "-il",
+            stdin=slave_fd, stdout=slave_fd, stderr=slave_fd,
+        )
+        os.close(slave_fd)
+        
+    def _start_reading(self):
+        """Register fd with asyncio event loop for event-driven reading."""
+        if self._read_handler_set or self._master_fd is None:
+            return
+        loop = asyncio.get_event_loop()
+        loop.add_reader(self._master_fd, self._on_fd_readable)
+        self._read_handler_set = True
+        
+    def _on_fd_readable(self):
+        """Callback when PTY fd has data available."""
+        if self._master_fd is None or self._paused:
+            return
+        try:
+            data = os.read(self._master_fd, 4096)
+            if data:
+                self._add_to_buffer(data)
+                self._queue_output(data)
+                self.last_activity = time.time()
+        except (OSError, IOError):
+            pass
+            
+    def _queue_output(self, data: bytes):
+        """Add to batch buffer, schedule flush."""
+        self._batch_buffer.extend(data)
+        self._unacknowledged_bytes += len(data)
+        
+        if self._unacknowledged_bytes > self.FLOW_CONTROL_THRESHOLD:
+            self._pause_output()
+            
+        if self._batch_timer is None:
+            loop = asyncio.get_event_loop()
+            self._batch_timer = loop.call_later(
+                self.BATCH_WINDOW_MS / 1000, self._flush_batch
+            )
+            
+    def _flush_batch(self):
+        """Flush batched output to all WebSockets."""
+        self._batch_timer = None
+        if not self._batch_buffer:
+            return
+            
+        payload = bytes(self._batch_buffer)
+        self._batch_buffer.clear()
+        
+        dead = set()
+        for ws in self._websockets:
+            try:
+                asyncio.create_task(ws.send_bytes(payload))
+            except Exception:
+                dead.add(ws)
+        
+        self._websockets -= dead
+        
+    def acknowledge_data(self, char_count: int):
+        """Client acknowledges processed bytes."""
+        self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
+        if self._paused and self._unacknowledged_bytes < self.FLOW_CONTROL_RESUME:
+            self._resume_output()
+            
+    def _pause_output(self):
+        """Pause reading from PTY."""
+        if self._read_handler_set and self._master_fd is not None:
+            loop = asyncio.get_event_loop()
+            loop.remove_reader(self._master_fd)
+            self._read_handler_set = False
+        self._paused = True
+        
+    def _resume_output(self):
+        """Resume reading from PTY."""
+        self._paused = False
+        self._start_reading()
+```
+
+## Frontend Changes
+
+### WebSocket Binary Mode
+```typescript
+const ws = new WebSocket(wsUrl);
+ws.binaryType = "arraybuffer";  // Receive ArrayBuffer directly
+
+ws.onmessage = (event) => {
+    if (event.data instanceof ArrayBuffer) {
+        const data = new Uint8Array(event.data);
+        termRef.current?.write(data);
+        
+        // Flow control: acknowledge processed bytes
+        ackAccumulator += data.length;
+        if (ackAccumulator >= 4096) {
+            ws.send(JSON.stringify({ type: "ack", chars: ackAccumulator }));
+            ackAccumulator = 0;
+        }
+    }
+};
+```
+
+### WebGL Renderer
+```typescript
+import { WebglAddon } from "xterm-addon-webgl";
+
+const webglAddon = new WebglAddon();
+try {
+    term.loadAddon(webglAddon);
+} catch (e) {
+    console.warn("WebGL failed, using DOM renderer", e);
+}
+```
+
+## WebSocket Protocol
+
+### Message Types
+
+**Client → Server:**
+- `{"type": "input", "data": "base64_encoded"}` — Keystrokes
+- `{"type": "resize", "cols": 80, "rows": 24}` — Resize
+- `{"type": "ack", "chars": 4096}` — Flow control acknowledgment
+- `{"type": "reset"}` — Reset session
+
+**Server → Client:**
+- Binary frame: raw terminal output bytes
+- `{"type": "status", "status": "connected"}` — Connection status
+- `{"type": "ping"}` — Heartbeat (server → client)
+
+## Testing
+
+### Unit Tests
+- `test_terminal_session_v2.py` — Test batching, flow control, resize, reset
+- `test_terminal_manager.py` — Test session lifecycle with new session class
+
+### Integration Tests
+- `test_terminal_websocket.py` — Full WebSocket round-trip
+
+### Performance Tests
+- `benchmark_terminal_latency.py` — Measure input/output latency
+- `benchmark_terminal_throughput.py` — Measure max throughput
+
+## Rollback Plan
+
+Since this is a full rewrite with no legacy support:
+- Keep a backup branch of the old terminal code
+- Feature flag in frontend: `?terminal=v2` to test before full rollout
+- Monitor error rates after deployment

openspec/explorations/terminal-responsiveness.md

@@ -0,0 +1,160 @@
+# SDD Exploration: Responsive Web Terminal
+
+## Status
+**Phase:** explore  
+**Date:** 2026-06-02  
+**Owner:** el Gentleman (parent session)  
+**Scope:** Terminal I/O latency, rendering performance, connection stability
+
+## Goal
+Achieve VS Code Server-level terminal responsiveness: near-local latency on keystrokes, smooth scrolling, no jank on output bursts, and instant resize reactions.
+
+## Current Architecture
+
+### Data Flow
+```
+Container shell → docker exec PTY → host PTY master fd → select.select(0.1s)
+  → Python read loop (10ms sleep fallback) → WebSocket.send_bytes()
+  → WebSocket (Blob mode) → frontend arrayBuffer decode → xterm.js.write()
+```
+
+### Key Files
+| File | Responsibility |
+|------|---------------|
+| `apps/web/src/components/terminal.tsx` | xterm.js, WebSocket client, FitAddon |
+| `apps/api/src/api/terminal.py` | WebSocket endpoint, auth, read/write/heartbeat loops |
+| `apps/api/src/services/terminal_session.py` | PTY creation, docker exec subprocess, I/O |
+| `apps/api/src/services/terminal_manager.py` | Session lifecycle, persistence, idle cleanup |
+
+### Current Bottlenecks
+
+#### 1. Blocking Read with 100ms Timeout
+```python
+# terminal_session.py:read_output()
+readable, _, _ = select.select([self._master_fd], [], [], 0.1)
+if readable:
+    data = os.read(self._master_fd, 4096)
+```
+**Problem:** `select.select` blocks up to 100ms when no data is available. With the read loop in `terminal.py` doing `asyncio.sleep(0.01)` between calls, worst-case latency from shell output to WebSocket is ~110ms.
+
+**VS Code approach:** node-pty uses libuv's epoll/kqueue watchers — event-driven, no polling timeout.
+
+#### 2. WebSocket Blob → arrayBuffer Conversion
+```typescript
+// terminal.tsx
+ws.onmessage = (event) => {
+    if (event.data instanceof Blob) {
+        event.data.arrayBuffer().then((buffer) => {
+            const data = new Uint8Array(buffer);
+            termRef.current?.write(data);
+        });
+    }
+}
+```
+**Problem:** Blob → arrayBuffer is async and adds microtask latency. Also forces GC pressure from transient Blobs.
+
+**VS Code approach:** Uses `ws.binaryType = "arraybuffer"` — receives ArrayBuffer directly, zero-copy into Uint8Array.
+
+#### 3. No asyncio-Native PTY Reading
+The PTY master fd is read with synchronous `os.read()` inside an async coroutine. This blocks the event loop thread for the duration of the read.
+
+**VS Code approach:** node-pty's C++ binding hooks into libuv's event loop natively — true async I/O.
+
+#### 4. Docker Exec Subprocess Overhead
+```python
+# terminal_session.py:start()
+self.process = await asyncio.create_subprocess_exec(
+    "docker", "exec", "-it", "-e", "TERM=xterm",
+    self.container_id, "bash", "-c", shell_cmd,
+    stdin=self._slave_fd, stdout=self._slave_fd, stderr=self._slave_fd,
+)
+```
+**Problem:** Spawns a new `docker exec` process on the host. Adds process startup latency and an extra process hop.
+
+**Alternative:** Docker Engine API's `attach` endpoint with `logs=0&stream=1&stdin=1&stdout=1&stderr=1` — streams directly to the API container via Unix socket. No host subprocess.
+
+#### 5. No Flow Control / Backpressure
+If a command dumps output faster than the WebSocket can send (e.g., `cat /dev/urandom | base64`), data piles up in:
+- The PTY kernel buffer (limited, ~4KB)
+- Python's deque circular buffer (10KB)
+- WebSocket's internal buffer (unbounded in some implementations)
+- xterm.js parser queue
+
+**VS Code approach:** Implements explicit flow control — pauses the PTY when the client buffer exceeds a threshold, resumes when drained.
+
+#### 6. xterm.js Renderer
+Current: DOM renderer (default).  
+**VS Code approach:** Canvas renderer with WebGL addon for GPU-accelerated rendering.
+
+## Measurement Baseline
+
+Before optimization, we need metrics:
+
+| Metric | How to Measure | Target |
+|--------|---------------|--------|
+| Input latency | Time from keypress to character appearing | < 16ms (1 frame) |
+| Output throughput | Bytes/sec for `cat /dev/zero` | > 1 MB/s |
+| Resize latency | Time from resize message to shell reacting | < 50ms |
+| Reconnection time | Time from disconnect to full replay | < 200ms |
+| Frame drops | Dropped frames during `yes` command | 0 |
+
+## Improvement Directions
+
+### Direction A: Low-Latency Read Loop (Quick Win)
+Replace `select.select` + `os.read` with `asyncio` native approach:
+- Use `loop.add_reader()` to register a callback when fd is readable
+- Or use `asyncio.to_thread()` with blocking `os.read` and immediate wake
+- Eliminate the 100ms timeout and 10ms sleep
+
+### Direction B: WebSocket Binary Mode (Quick Win)
+Set `ws.binaryType = "arraybuffer"` on frontend, send binary frames directly.
+Eliminates Blob → arrayBuffer conversion.
+
+### Direction C: Docker Engine API Attach (Medium)
+Replace `docker exec` subprocess with direct container attach via Docker SDK or HTTP API:
+```python
+from docker import DockerClient
+client = DockerClient()
+container = client.containers.get(container_id)
+socket = container.attach_socket(params={...})
+# socket is a raw TCP/Unix socket — read with asyncio
+```
+**Pros:** No subprocess overhead, direct stream to container  
+**Cons:** Requires Docker SDK or raw HTTP over Unix socket; needs `docker` group permissions
+
+### Direction D: Flow Control (Medium)
+Add backpressure mechanism:
+1. Measure WebSocket send buffer depth on backend
+2. Pause reading from PTY when buffer exceeds threshold (e.g., 64KB)
+3. Resume when buffer drains below threshold
+4. Frontend: measure xterm.js parser queue depth, pause via control message
+
+### Direction E: WebGL Renderer (Quick Win)
+Add xterm-addon-webgl:
+```typescript
+import { WebglAddon } from 'xterm-addon-webgl';
+term.loadAddon(new WebglAddon());
+```
+**Pros:** GPU-accelerated, much faster for large output bursts  
+**Cons:** Falls back to canvas/DOM if WebGL unavailable; slightly higher init time
+
+### Direction F: Output Batching (Quick Win)
+Batch small writes before sending over WebSocket:
+- Collect output for 1-2ms
+- Send as single binary frame
+- Reduces WebSocket frame overhead for high-frequency small writes (e.g., progress bars)
+
+## Recommended Next Steps
+
+1. **Measure baseline** with synthetic benchmarks
+2. **Implement Directions A + B + F** (low-risk, high-impact)
+3. **Evaluate Direction C** (Docker API attach) vs keeping docker exec
+4. **Add Direction D** (flow control) if throughput tests show issues
+5. **Add Direction E** (WebGL) as frontend enhancement
+
+## Risks
+
+- Docker API attach may not support PTY mode as cleanly as `docker exec -it`
+- WebGL addon may have compatibility issues on older GPUs
+- Flow control adds complexity; premature optimization risk
+- Changes to core I/O loop could introduce stability regressions

openspec/proposals/terminal-responsiveness.md

@@ -0,0 +1,71 @@
+# Proposal: High-Performance Web Terminal
+
+## Status
+**Phase:** proposal → spec → design → tasks → apply  
+**Date:** 2026-06-02  
+**Owner:** el Gentleman  
+**Scope:** Terminal I/O latency, rendering performance, connection stability
+
+## Problem
+
+The current web terminal has noticeable latency on keystrokes, choppy scrolling, and poor performance during output bursts. Users report it feels "slow" compared to VS Code Server's terminal, which feels almost local.
+
+## Goals
+
+| Metric | Current | Target | How Measured |
+|--------|---------|--------|--------------|
+| Input latency (keypress → char visible) | ~110ms | < 16ms (1 frame) | `term.write()` timestamp diff |
+| Output throughput (`cat /dev/zero`) | ~200KB/s | > 1 MB/s | Bytes/sec over 5s |
+| Resize latency | ~200ms | < 50ms | Time from resize msg to shell SIGWINCH |
+| Reconnection + replay | ~2s | < 300ms | Time from WS open to first rendered char |
+| Frame drops during `yes` | Many | 0 | `requestAnimationFrame` counter |
+
+## Non-Goals
+
+- Changing the terminal UI/UX (chrome, controls, tabs)
+- Adding new terminal features (search, multi-cursor, etc.)
+- Changing authentication or session persistence model
+- Supporting non-Docker container runtimes
+
+## Constraints
+
+- Must work with existing tool instance lifecycle (docker containers)
+- Must preserve WebSocket-based architecture
+- Must preserve session persistence across reconnections
+- Must work in both development and production compose setups
+
+## Solution Overview
+
+Replace the blocking `select.select()` PTY read loop with asyncio-native event-driven I/O. Replace `docker exec` subprocess with Docker Engine API attach. Switch WebSocket to binary mode. Add output batching. Add WebGL renderer.
+
+## Key Decisions
+
+1. **Keep `docker exec` for now** — Docker SDK attach doesn't support PTY mode as cleanly. We can optimize the subprocess approach with proper fd handling.
+2. **Use `asyncio.add_reader()`** — Native asyncio event-driven fd reading eliminates polling latency.
+3. **Binary WebSocket frames** — `ws.binaryType = "arraybuffer"` eliminates Blob conversion overhead.
+4. **WebGL renderer with DOM fallback** — GPU acceleration where available, graceful fallback.
+5. **Output batching with 2ms window** — Collect small writes before sending to reduce frame overhead.
+6. **Flow control v2** — Client acknowledges processed bytes; server pauses reads when buffer is full.
+
+## Risks
+
+- **Event loop blocking**: `asyncio.add_reader()` on a PTY fd may not work on all platforms (should work on Linux)
+- **WebGL compatibility**: Some GPUs/drivers may fail WebGL context creation
+- **Docker exec subprocess**: Still adds overhead; may revisit Docker API attach in future
+- **Full rewrite**: Large change surface; thorough testing required
+
+## Acceptance Criteria
+
+- [ ] Input latency < 16ms measured with synthetic benchmark
+- [ ] Output throughput > 1 MB/s measured with `cat /dev/zero`
+- [ ] Resize latency < 50ms
+- [ ] Reconnection + replay < 300ms
+- [ ] No frame drops during `yes` command
+- [ ] All existing terminal tests pass
+- [ ] WebGL renderer loads successfully on modern browsers
+- [ ] Graceful fallback to DOM renderer if WebGL fails
+- [ ] Flow control prevents memory bloat on `cat /dev/urandom | base64`
+
+## Related
+
+- `openspec/explorations/terminal-responsiveness.md` — Detailed bottleneck analysis

openspec/specs/terminal-responsiveness.md

@@ -0,0 +1,176 @@
+# Spec: High-Performance Web Terminal
+
+## Overview
+
+Complete rewrite of the terminal I/O pipeline for sub-frame latency and smooth rendering.
+
+## Architecture
+
+### Data Flow (New)
+
+```
+Container shell → docker exec PTY → host PTY master fd
+  → asyncio.add_reader() callback (event-driven, zero polling)
+    → output batcher (2ms window) → WebSocket.send_bytes()
+      → WebSocket binary frame → frontend ArrayBuffer
+        → xterm.js WebGL renderer → screen
+```
+
+### Components
+
+#### 1. TerminalSession (backend)
+
+**Responsibilities:**
+- Create PTY via `pty.openpty()`
+- Spawn `docker exec -it` with slave fd attached
+- Read from PTY master fd using `asyncio.add_reader()`
+- Batch output (2ms window) before sending to WebSocket
+- Handle flow control (pause/resume reads based on client ack)
+- Resize via `TIOCSWINSZ` + `SIGWINCH`
+
+**Interface:**
+```python
+class TerminalSession:
+    async def start(self) -> None
+    async def read_loop(self, websocket) -> None  # event-driven
+    async def write_input(self, data: bytes) -> None
+    async def resize(self, cols: int, rows: int) -> None
+    async def reset(self) -> None
+    async def close(self) -> None
+    
+    # Flow control
+    def acknowledge_data(self, char_count: int) -> None
+    def pause_output(self) -> None
+    def resume_output(self) -> None
+```
+
+#### 2. TerminalManager (backend)
+
+Unchanged responsibilities (session lifecycle, persistence, idle cleanup).
+
+#### 3. WebSocket Handler (backend)
+
+**Messages:**
+
+| Direction | Type | Payload | Description |
+|-----------|------|---------|-------------|
+| C → S | `input` | `{"data": "base64"}` | Keystrokes / input |
+| C → S | `resize` | `{"cols": 80, "rows": 24}` | Terminal resize |
+| C → S | `ack` | `{"chars": 1024}` | Flow control ack |
+| C → S | `reset` | `{}` | Reset session |
+| S → C | `binary` | raw bytes | Terminal output |
+| S → C | `status` | `{"status": "connected"}` | Connection status |
+| S → C | `ping` | `{}` | Heartbeat |
+
+**Key changes:**
+- Output is sent as **binary WebSocket frames**, not Blob
+- Flow control: server tracks unacknowledged bytes, pauses PTY reads at 64KB threshold
+
+#### 4. TerminalComponent (frontend)
+
+**Key changes:**
+- `ws.binaryType = "arraybuffer"` before connection
+- Binary frames written directly to xterm.js as `Uint8Array`
+- Flow control: send `ack` messages every 4096 processed bytes
+- WebGL renderer with DOM fallback
+- Batch resize messages (debounce 50ms)
+
+**xterm.js config:**
+```typescript
+const term = new Terminal({
+    cursorBlink: true,
+    fontSize: currentFontSize,
+    fontFamily: 'Menlo, Monaco, "Courier New", monospace',
+    lineHeight: 1.2,
+    letterSpacing: 0,
+    allowTransparency: false,
+    scrollback: 10000,
+    // Performance options
+    ignoreBracketedPasteMode: false,
+    fastScrollSensitivity: 5,
+    scrollSensitivity: 1,
+});
+```
+
+### Flow Control Protocol
+
+**Server-side buffer tracking:**
+```python
+self._unacknowledged_bytes = 0
+self._flow_control_threshold = 64 * 1024  # 64KB
+self._paused = False
+
+def on_output(self, data: bytes) -> None:
+    self._unacknowledged_bytes += len(data)
+    if self._unacknowledged_bytes > self._flow_control_threshold:
+        self.pause_output()
+    
+def acknowledge_data(self, char_count: int) -> None:
+    self._unacknowledged_bytes = max(0, self._unacknowledged_bytes - char_count)
+    if self._paused and self._unacknowledged_bytes < self._flow_control_threshold / 2:
+        self.resume_output()
+```
+
+**Client-side ack strategy:**
+- After every `term.write(data)`, accumulate processed bytes
+- Send `ack` message every 4096 bytes or 100ms
+
+### Output Batching
+
+**Server-side batcher:**
+```python
+self._batch_buffer = bytearray()
+self._batch_timer: asyncio.TimerHandle | None = None
+self._batch_window_ms = 2
+
+def queue_output(self, data: bytes) -> None:
+    self._batch_buffer.extend(data)
+    if self._batch_timer is None:
+        self._batch_timer = asyncio.get_event_loop().call_later(
+            self._batch_window_ms / 1000, self._flush_batch
+        )
+
+async def _flush_batch(self) -> None:
+    self._batch_timer = None
+    if self._batch_buffer and websocket.open:
+        await websocket.send_bytes(bytes(self._batch_buffer))
+        self._batch_buffer.clear()
+```
+
+## Docker Exec Subprocess
+
+**Command:**
+```bash
+docker exec -it -e TERM=xterm-256color <container_id> bash -il
+```
+
+**Why keep docker exec:**
+- Docker SDK `attach()` doesn't support PTY allocation cleanly
+- `docker exec -it` is the standard way to get an interactive TTY
+- Subprocess overhead is acceptable compared to PTY latency improvements
+
+**Optimization:** Pre-warm the connection by reusing the same `docker exec` process for the session lifetime.
+
+## Error Handling
+
+| Scenario | Behavior |
+|----------|----------|
+| WebGL init fails | Fall back to DOM renderer, log warning |
+| Flow control ack lost | Server resumes after timeout (5s) |
+| PTY fd closed | Close WebSocket with code 4004 |
+| Docker exec exits | Close WebSocket with code 4004, allow reconnect |
+| Binary frame too large | Split into multiple frames (max 64KB) |
+
+## Testing Strategy
+
+1. **Unit tests:** Mock PTY fd, verify batching, flow control, resize
+2. **Integration tests:** Full WebSocket round-trip with test container
+3. **Performance tests:**
+   - `time cat /dev/zero | head -c 10M` — measure throughput
+   - Rapid keypress script — measure input latency
+   - Resize storm — measure resize latency
+4. **Browser tests:** WebGL fallback on devices without GPU
+
+## Migration
+
+Full rewrite — no migration needed. Old terminal code can be deleted.

openspec/tasks/terminal-responsiveness.md

@@ -0,0 +1,60 @@
+# Tasks: High-Performance Web Terminal
+
+## Task 1: Rewrite TerminalSession with asyncio-native I/O
+**Status:** pending
+**Files:** `apps/api/src/services/terminal_session.py` (full rewrite)
+**Description:**
+- Replace `select.select()` with `asyncio.add_reader()` for event-driven PTY reading
+- Add output batching (2ms window)
+- Add flow control (pause/resume based on client ack)
+- Keep docker exec subprocess (optimized)
+- Remove circular buffer (not needed with event-driven architecture)
+
+## Task 2: Update TerminalManager for new session class
+**Status:** pending
+**Files:** `apps/api/src/services/terminal_manager.py`
+**Description:**
+- Update imports to use rewritten TerminalSession
+- Verify session lifecycle methods still work
+- Update DB persistence calls
+
+## Task 3: Update WebSocket endpoint for binary frames + flow control
+**Status:** pending
+**Files:** `apps/api/src/api/terminal.py`
+**Description:**
+- Accept binary output frames from TerminalSession
+- Handle `ack` flow control messages from client
+- Send `ping` heartbeat
+- Maintain existing auth and session management
+
+## Task 4: Update frontend for binary WebSocket + WebGL
+**Status:** pending
+**Files:** `apps/web/src/components/terminal.tsx`, `apps/web/package.json`
+**Description:**
+- Set `ws.binaryType = "arraybuffer"`
+- Send flow control `ack` messages
+- Add xterm-addon-webgl with DOM fallback
+- Optimize resize handling
+
+## Task 5: Add performance benchmarks
+**Status:** pending
+**Files:** `apps/api/tests/benchmark_terminal.py`
+**Description:**
+- Input latency benchmark
+- Output throughput benchmark
+- Resize latency benchmark
+- Reconnection time benchmark
+
+## Task 6: Update/fix unit tests
+**Status:** pending
+**Files:** `apps/api/tests/unit/test_tool_instances_legacy.py`, new tests
+**Description:**
+- Fix any tests broken by terminal changes
+- Add tests for new TerminalSession features
+
+## Task 7: Run full test suite
+**Status:** pending
+**Description:**
+- Run all API tests
+- Verify no regressions
+- Report quality gate results