The Jules MCP Server exposes Google Jules Agent operations via FastMCP, allowing MCP clients to discover Jules sources, manage sessions, and inspect activities through the jules-agent-sdk. The server organizes capabilities into three tool areas: Sources, Sessions, and Activities. You can query sources with list_sources or get_all_sources, create and manage sessions with create_session, approve_session_plan, send_session_message, and wait_for_session_completion, and inspect or enumerate activities with get_activity, list_activities, or list_all_activities. The official signatures and inline docs live in the jules_mcp/jules_mcp.py module, which you can reference for parameter shapes and return structures. When running programmatically, you can interact through the FastMCP client, calling tools by name with the appropriate argument payloads.

To use the server from a Python environment, you typically either run the module as a standalone server or instantiate a FastMCP Client against the running server. The examples show how to list sources, create a session with a prompt and source, optionally approve a plan, wait for completion, and then inspect activities. Environment configuration, notably the Jules API key, must be provided so the SDK can authenticate API requests to Jules.

Prerequisites:

• Python 3.13+ installed on your system
• Access to the Jules API and a valid API key (JULES_API_KEY)
• Optional: uv (for development) and a Python virtual environment tool

Installation steps:

1. Create and activate a virtual environment (recommended): python -m venv .venv

   macOS/Linux

   source .venv/bin/activate

   Windows (PowerShell)

   ..venv\Scripts\Activate.ps1

2. Install the package in editable mode (from local checkout): pip install -e .

3. Install development dependencies (optional): pip install -e .[dev]

4. (Optional) Install uv for development workflow as described in the README: pip install uv

5. Ensure your environment variable for Jules API key is set prior to running the server:

   • Unix shells: export JULES_API_KEY="<your_api_key_here>"
   • Windows PowerShell: $Env:JULES_API_KEY = "<your_api_key_here>"

6. Run the server (module entry point): python -m jules_mcp

If you prefer running via uv for development, see the README’s guidance on using uv to run the MCP server.

Tips and notes:

• The server expects a Jules API key; without it, many operations will fail. Configure JULES_API_KEY in your environment.
• The mcp config supports running the server via the Python module entry point or as a standalone FastMCP server. The README shows both in-process usage and external server execution.
• When using the Python client examples, ensure you are connected to the same MCP instance and that the mcp object is correctly initialized in jules_mcp.
• For debugging, run with development dependencies installed and consider running tests with uv run pytest -q, noting that Jules API calls may require a valid key.
• If you modify paths or checkout locations, adjust any MCP.json or fastmcp.json configuration references accordingly.