Aider MCP Server - Experimental

What is Aider MCP Server - Experimental?

A Model Context Protocol server that allows Claude Code to offload AI coding tasks to Aider, the open source AI coding assistant. It helps reduce costs, gives control over the coding model, and allows Claude Code to operate orchestratively by reviewing and revising code.

How to use Aider MCP Server - Experimental?

Clone the repository, run uv sync, configure API keys in .env, then add the server to Claud Code via claude mcp add with the appropriate --editor-model and --current-working-dir. Once running, use the aider_ai_code tool with a prompt and editable files, or list_models to discover available models.

Key features of Aider MCP Server - Experimental

• Offloads AI coding tasks from Claude Code to Aider.
• Uses the aider_ai_code tool with prompt and file paths.
• Lists available models matching a substring via list_models.
• Supports multiple AI model providers (OpenAI, Anthropic, Gemini, etc.).
• Configurable editor model for separate coding and editing.
• Operates over stdio transport with environment-based API keys.

Use cases of Aider MCP Server - Experimental

• Refactoring a function in a Python file to handle exceptions.
• Creating a new file based on natural language instructions.
• Discovering which models are available for a given substring.
• Using a cheaper or more capable model for editing than for initial coding.

FAQ from Aider MCP Server - Experimental

What dependencies are required?

Python, uv, and API keys for the chosen AI model (set in .env or the env section of mcpServers).

How do I specify which model Aider should use?

Use the --editor-model argument when adding the server, and optionally the model parameter in aider_ai_code calls.

Where are my project files located during execution?

All file paths are relative to the --current-working-dir specified in the server configuration.

What transport does the server use?

It uses stdio transport, as defined in the .mcp.json configuration.

Can I test the server locally?

Yes, tests are run with uv run pytest and require a valid Gemini API key for AI coding tests.