This project provides an experimental Model Context Protocol (MCP) server for Aider, an open-source AI coding assistant. It enables AI coding assistants like Claude Code to offload AI coding tasks to Aider, offering cost reduction, model control, and enhanced orchestration for code review and revision.

How It Works

The server acts as an intermediary, exposing Aider's capabilities as MCP tools. It leverages Aider's core functionality to perform AI coding tasks based on user prompts and specified files, returning success status and code diffs. It also provides a tool to list available AI models compatible with Aider, facilitating model discovery. The architecture is layered, separating MCP protocol handling, core functional components (atoms), specific tool implementations, utilities, and data types.

Quick Start & Requirements

• Install dependencies: uv sync
• Configure API keys (Gemini, OpenAI, Anthropic) in .env.
• Add to Claude Code using claude mcp add ... commands, specifying the Aider MCP server path and desired editor model.
• Requires Python and uv.

Highlighted Details

• Offloads AI coding tasks to Aider, supporting specified editable and read-only files.
• Provides a list_models tool to discover Aider-compatible AI models.
• Supports multiple AI models including Gemini, GPT-4o, and others via OpenRouter or Fireworks AI.
• Tested with Gemini models; AI coding tests require a valid Gemini API key.

Maintenance & Community

No specific community channels or notable contributors are mentioned in the README.

Licensing & Compatibility

The repository does not explicitly state a license.

Limitations & Caveats

The project is described as "Experimental." The README does not specify compatibility with commercial or closed-source applications. A license is not explicitly stated, which may impact adoption.