Aider MCP Server

What is Aider MCP Server?

A Machine Cognition Protocol (MCP) server that provides AI coding capabilities using Aider. It leverages Aider, a powerful AI coding assistant, to offer coding via a standardized API, reducing costs by discretely offloading work to Aider while using model-specific capabilities for more reliable code through multiple focused LLM calls.

How to use Aider MCP Server?

Install by cloning the repository and running uv pip install -e . in a virtual environment. Configure the server via a .env file (setting transport, host, port, and API keys) or CLI options (--editor-model, --architect-model, --cwd). Start in SSE mode with uv run python -m aider_mcp_server, or configure an MCP client to run it automatically via stdio or Docker.

Key features of Aider MCP Server

• AI code generation for one-shot coding tasks
• Model selection to query available Aider models
• Flexible configuration with customizable Aider session settings
• Multi-transport support via SSE or stdio

Use cases of Aider MCP Server

• Add a new function or feature to an existing codebase
• Fix bugs or enhance code with a targeted AI prompt
• Query available AI models to choose the best option for a task

FAQ from Aider MCP Server

What tools does the server expose?

The server exposes ai_code for performing coding tasks and get_models for listing available Aider models filtered by a substring.

What transports are supported?

The server supports both Server-Sent Events (SSE) and stdio transport modes.

What are the Python and package manager requirements?

Python 3.10 or higher is required. The recommended package manager is uv, though pip can also be used.

What API keys are needed to run the server?

API keys from OpenAI, Anthropic, or Google (Gemini) are required depending on which models you intend to use. Each key is configured via environment variables (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY).

How do I configure the server's transport and network settings?

Set the TRANSPORT environment variable (default sse), and optionally HOST (default 0.0.0.0) and PORT (default 8050) for SSE mode. The editor model can be set via --editor-model (default gemini/gemini-2.5-pro-exp-03-25).