Arbor

You get a dedicated MCP (Model Context Protocol) server that enables your AI tools to manage and reuse context across sessions. It provides commands to log reasoning, create and switch context branches, track progress, and merge context, so you can stay organized while your AI helps you work through tasks.

How to use

You will connect an MCP client to the server and use its tools to manage your project context. The server exposes a set of commands that your AI or CLI can invoke to log reasoning, commit checkpoints, branch into new tasks, merge work from multiple branches, and query project information. When you start a chat or a new session, the MCP server can auto-detect your workspace and begin organizing your context without manual steps.

Core capabilities you can leverage through the MCP server include:

• LOG: Record reasoning steps as the AI works through problems
• COMMIT: Create checkpoints at meaningful progress points
• BRANCH: Create or switch to task-specific branches with access to prior context
• MERGE: Combine context from multiple branches into the current view
• INFO: Retrieve project, branch, and session information at various levels
• TODO Management: Track tasks across branches and sessions
• Smart Branch Detection: Automatically identify the appropriate branch for your current work

How to install

Before you begin, ensure Python 3.8+ is installed on your system and that you have network access to install dependencies.

Step 1: Install Python dependencies and set up the MCP server package

pip install -r requirements.txt
# Or install in development mode:
pip install -e .

Step 2: Create or verify the MCP server configuration for your MCP client (example uses the stdio-based local server)

{
  "mcpServers": {
    "context_mcp": {
      "type": "stdio",
      "command": "python3",
      "args": ["-m", "src.mcp_server"]
    }
  }
}

Additional configuration notes

No manual changes are required after the server is running; the MCP client will handle workspace detection and context directory creation automatically when the server starts.

Troubleshooting and tips

If the MCP server does not respond, check that the Python process starts correctly and that the client is pointing to the right command and module. Ensure dependencies are installed and that the working directory is accessible.

Project structure and persistence

The server stores context data in a dedicated directory within your project workspace. It creates branches, commits histories, and logs that persist across sessions so you can resume work where you left off.

What you get after setup

Automatic workspace detection, automatic branching for different tasks, continuous logging of reasoning steps, checkpointing progress, and seamless return to prior work across sessions.

Available tools

context_status

Check workspace path and current branch. Used at session start.

context_set_workspace

Set the project workspace directory if auto-detection fails.

context_summary

Get quick overview of TODOs and progress. Used at session start and when checking status.

context_detect_branch

Smart branching: find the right branch for current work. Used before starting new tasks.

context_todos

View, add, or complete TODO items. Used throughout the session.

context_log

Log a reasoning step during AI thought process.

context_commit

Checkpoint progress by creating a commit.

context_branch

Create a new branch or switch to an existing one.

context_merge

Merge context from other branches.

context_info

Get detailed project/branch/session information.