Reachy Claude

Reachy Claude MCP lets Reachy Mini respond to your coding sessions with emotions, speech, and playful dances, creating an engaging, interactive coding experience. You can run Reachy in simulation or with a physical robot, connect it to Claude Code, and use a rich set of tools to get real-time feedback, memory, and smart responses during development.

How to use

You will run the MCP server for Reachy Claude and connect it to your Claude Code environment. Start the Reachy Claude MCP server locally, ensure Reachy or the simulation is available, and then configure Claude Code to communicate with the server. During your coding sessions, you can invoke tools to have Reachy speak, show emotions, celebrate milestones, or search past solutions. The system adapts based on whether you enable LLM features or memory, so you can tailor the experience to your needs.

How to install

Prerequisites: Python 3.10+ and optionally Docker for the vector database. You will also need either a Reachy Mini robot or the simulation to test interactions, and audio output for speech playback.

Additional setup and configuration

Configure how Claude Code talks to Reachy Claude MCP by adding the MCP server entry to your client configuration. The server can run as a local process via a stdio interface.

Troubleshooting

If you encounter audio issues, ensure Piper TTS voice data is available or configured, and confirm your audio output device is active. For LLM backends, verify that the selected backend (MLX on Apple Silicon or Ollama) is accessible, or fall back to keyword-only sentiment if no backend is available.

Notes on usage

When you start the simulation or connect to a real Reachy Mini, Reachy can greet you at session start, react to code changes, and celebrate milestones. Use the available tools to drive conversations, animations, and responses that align with your coding workflow.

Configuration

Environment variables control the behavior and integration points for Reachy Claude MCP. The key variables cover home directory, language models, and memory/vector store settings.

Tools you can use during coding

During sessions, you can call a range of tools to control Reachy's behavior, speak, and react to your progress. Examples include issuing greetings, starting and stopping interactions, or triggering expressive responses.

Security and access

Keep any API-like credentials or tokens secure. Use environment variables to configure access to local services (LLM backends, memory stores) and avoid hard-coding sensitive values in scripts.

Migration and compatibility

If you enable memory features or enhanced LLM capabilities, you may need to start supporting services like Qdrant or Ollama and adjust environment variables accordingly. Ensure your platform matches the supported configurations for the selected features.

Installation options and commands

To install the Reachy Claude MCP package and prepare to run the server, follow the steps below.

Step-by-step install commands

pip install reachy-claude-mcp

Start the Reachy Mini simulation (if you don’t have a physical robot)

On macOS with Apple Silicon use the MuJoCo-based Python runner. For macOS Apple Silicon: run the simulation as shown. For other platforms, use the standard Python invocation.

Start simulation (example commands)

# macOS with Apple Silicon
mjpython -m reachy_mini.daemon.app.main --sim --scene minimal

# Other platforms (Linux/Windows/Intel Mac)
python -m reachy_mini.daemon.app.main --sim --scene minimal

Configure the MCP client

Add the Reachy Claude MCP server to your Claude Code configuration so the client knows how to reach the server.

Example MCP configuration in Claude Code

{
  "mcpServers": {
    "reachy_claude": {
      "command": "reachy-claude"
    }
  }
}

Available tools

robot_respond

Speak a summary of the current state or result and play an emotion animation

robot_emotion

Play a specific emotion animation without speaking

robot_celebrate

Perform a celebration animation and announce success

robot_thinking

Show a thinking/processing animation during long operations

robot_wake_up

Greet the user at the start of a session

robot_sleep

Say goodbye and end the session

robot_dance

Perform a dance move to express feedback

robot_big_celebration

Celebrate a major milestone with a big reaction