aider

The Aider MCP Server exposes AI coding capabilities via a standardized MCP interface. It uses Aider to perform coding tasks, model selection, and general question answering, accessible through tools such as ai_code for coding tasks, get_models to list available models, and ask_question for direct prompts to an LLM. The server supports multiple transports (SSE or stdio) to fit into your integration workflow, and lets you configure Aider sessions with a range of environment variables, including API keys for OpenAI, Anthropic, and Google Gemini. You can run the server in SSE mode to expose a streaming API endpoint, or use stdio mode where the MCP client starts the server automatically. The included MCP tools provide concrete JSON inputs and examples to help you orchestrate AI-assisted coding tasks or queries within your applications.

Prerequisites:

• Python 3.10 or higher
• A UDP/Internet-connected environment to install dependencies
• Access to an MCP client library or environment capable of HTTP/WS streaming or stdio integration

Install steps:

1. Clone the repository: git clone https://github.com/your-username/aider-mcp.git cd aider-mcp

2. Create and activate a Python virtual environment (recommended): python -m venv venv source venv/bin/activate # On Windows use: venv\Scripts\activate

3. Install the package in editable mode: uv pip install -e .

4. Run tests to verify setup: uv run pytest

5. Start the server (example for SSE mode): uv run python -m aider_mcp_server

Note: Ensure API keys for any models you intend to use (OpenAI, Anthropic, Gemini) are provided via environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY) as needed.

Environment variables and configuration:

• TRANSPORT: sse (default) or stdio when using the MCP client in stdio mode.
• HOST, PORT: Bind address and port for SSE transport as needed.
• API keys: Provide OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY only if you intend to use the corresponding models.

Common issues:

• If the server fails to start, verify your Python version and that the aider_mcp_server module is importable in your environment.
• When using stdio mode, the MCP client must be configured to spawn the server with the correct command and environment (e.g., TRANSPORT=stdio).
• Ensure network accessibility if SAP clients connect over SSE; firewall rules may block the SSE endpoint.

Tips:

• Use get_models to discover available models before running ai_code tasks.
• For code generation tasks, supply a clear ai_coding_prompt and restrict editable files to the project scope.
• When using Docker for deployment, mount your project path and pass API keys as environment variables.