aether-agent

Aether exposes its internal SQLite Brain via the MCP (Model Context Protocol) interface, allowing your IDE or other MCP-compatible clients to access its long-term memory and Morning Briefs. The server is driven by a Python FastAPI kernel on the backend and can be connected to via MCP clients such as Claude Desktop, Cursor, or VS Code integrations. By configuring an MCP entry (as shown in the example), your editor can query Aether’s memory, share context, and synchronize ongoing sessions with the agent’s world model.

To use the MCP bridge, add a server entry to your claude_desktop_config.json or Cursor settings that points to the aether-agent backend. Once the MCP server is running, you can connect to it from your IDE or editor, enabling seamless access to Aether’s concepts, memories, and daily briefings as you work on projects. This enables long-term memory sharing and cross-AI collaboration with FastMCP-compatible tooling.

Prerequisites

• Python 3.12+ and a functioning uv (Python) environment, or the uv toolchain as described in the project setup.
• uv Python manager installed (as per the repository setup instructions).
• Optional: Node.js 20+ and pnpm if you plan to run the frontend/desktop components locally.

Manual installation steps

1. Install UV (Python environment manager) if not already installed: curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install Node.js (if you intend to build or run the frontend):

   • Ensure Node.js 20+ is installed from nodejs.org
   • npm install -g pnpm

3. Prepare the backend environment

   • Ensure Python 3.12+ is available
   • Clone the repository and navigate to the backend directory

4. Install backend dependencies via uv and start the MCP server

   • Follow the project’s one-click setup or run the following equivalent commands: uv install --project /PATH/TO/aether-agent/backend uv run --project /PATH/TO/aether-agent/backend python /PATH/TO/aether-agent/backend/mcp_server.py

5. Run the MCP server configuration

   • Use the example mcp_config above to register the MCP server in your editor configuration.

6. Optional: Docker or Docker-Compose setup

   • If you prefer Docker, use the provided docker-compose.yml in the repository and follow the Docker-based deployment steps in the README.

Notes and tips:

• Replace /PATH/TO/aether-agent/backend with the actual absolute path to your local aether-agent backend directory when configuring the MCP server in your editor.
• Ensure your environment variables (e.g., API keys for Gemini/Tavily/Telegram) are set in the appropriate .env files if you plan to run the full Aether stack.
• The MCP configuration shown uses the 'uv' toolchain to run the Python backend; your setup may vary depending on your environment (paths, virtual environments, etc.).
• If you encounter connectivity issues, verify that the MCP client and server share compatible FastMCP settings and that the backend server is listening on the correct interface.
• For production-like use, consider enabling the provided Docker/Compose setup for consistent environments and memory persistence.