Worktree Control
An open-source scheduling and resource management tool for multiple AI coding agents, centrally managing development server ports, build and test queues, and worktree claiming based on the semaphore concept.
Timeline (last 90 days)
Worktree Control is an open-source tool for managing multiple AI coding assistant agents, offering dev server scheduling, test queue management and a shared knowledge base.
Worktree Control is released as open source on GitHub, with the source code at https://github.com/pioootrek/worktree-control.
Worktree Control's core design limits the number of managed dev servers running at once, assigns each project a stable port, and requires an agent to claim a project before use and release it when done.
Worktree Control's builds and tests go through separate queues that limit concurrent runs, with at most one active check per worktree.
Worktree Control records Git observations (HEAD hash, whether the worktree is dirty, a summary of changed paths, etc.) before and after running a command, keeping command results separate from source evidence.
Local checks use the worktree files (including uncommitted changes) rather than immutable snapshots, and Git observations cannot capture all environment dependencies.
Worktree Control includes a Knowledge module for storing to-dos, discussions and project memory; tasks can be linked to the originating discussion, and decisions retain their source and revision history.
Agents access Worktree Control's knowledge base via MCP or the CLI, and any coding client that supports MCP can access the shared context in a consistent way.
The Worktree Control controller is a single Node.js process using SQLite and a browser dashboard, with the Next.js UI exported as static files, requiring no additional always-on server.
The Worktree Control project is still in the pre-1.0 stage, and distributing verification to independent worker nodes is still under development.
2 more timeline events
All Facts (12)
Worktree Control is an open-source tool for managing multiple AI coding assistant agents, offering dev server scheduling, test queue management and a shared knowledge base.
50%UnverifiedWorktree Control is released as open source on GitHub, with the source code at https://github.com/pioootrek/worktree-control.
50%UnverifiedWorktree Control's core design limits the number of managed dev servers running at once, assigns each project a stable port, and requires an agent to claim a project before use and release it when done.
50%UnverifiedWorktree Control's builds and tests go through separate queues that limit concurrent runs, with at most one active check per worktree.
50%UnverifiedWorktree Control records Git observations (HEAD hash, whether the worktree is dirty, a summary of changed paths, etc.) before and after running a command, keeping command results separate from source evidence.
50%UnverifiedLocal checks use the worktree files (including uncommitted changes) rather than immutable snapshots, and Git observations cannot capture all environment dependencies.
50%UnverifiedWorktree Control includes a Knowledge module for storing to-dos, discussions and project memory; tasks can be linked to the originating discussion, and decisions retain their source and revision history.
50%UnverifiedAgents access Worktree Control's knowledge base via MCP or the CLI, and any coding client that supports MCP can access the shared context in a consistent way.
50%UnverifiedThe Worktree Control controller is a single Node.js process using SQLite and a browser dashboard, with the Next.js UI exported as static files, requiring no additional always-on server.
50%UnverifiedThe Worktree Control project is still in the pre-1.0 stage, and distributing verification to independent worker nodes is still under development.
50%UnverifiedWorktree Control can lock a project via the dashboard to prevent other agents from grabbing it; releasing a claim does not stop the server, so an idle reservation is not the same as an idle slot.
50%UnverifiedWorktree Control's scheduling rules only take effect when agents use the controller, and cannot prevent unrelated terminals from starting extra servers, so the workflow instructions are also written into the project agent instructions.
50%