Reachy Mini MCP

Reachy Mini MCP

MCP server for Pollen Robotics Reachy Mini robot control.

For AI systems - Token-efficient reference for programmatic use.

Quick Start

# Install
cd reachy-mini-mcp
poetry install

# Set TTS key (optional, for speak())
export DEEPGRAM_API_KEY=your_key_here

# Start simulator daemon
mjpython -m reachy_mini.daemon.app.main --sim --scene minimal

# Run MCP server
poetry run python src/server.py

Architecture

MCP Tool → SDK Call → Daemon → Robot/Simulator

High-level express() abstracts motor control into semantic actions.Low-level tools available for precise control when needed.

Tools

Expression (Preferred)

Built-in emotions (express): neutral, curious, uncertain, recognition, joy, thinking, listening, agreeing, disagreeing, sleepy, surprised, focused

Recorded moves (play_move): 40+ from Pollen's HuggingFace libraries - fear1, rage1, loving1, serenity1, contempt1, dance3, etc. Use list_moves() to discover all options.

Motor Control

I/O

Lifecycle

Expression Vocabulary

Each emotion maps to choreography:

• Head pose: roll, pitch, yaw (degrees)
• Antennas: left, right angles (degrees)
• Duration: movement time (seconds)
• Interpolation: linear, minjerk, ease_in_out, cartoon

Example: curious → head forward (pitch +10, yaw +8), antennas up (+20, +20), 1.2s, ease_in_out

MCP Config

Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "reachy-mini": {
      "command": "poetry",
      "args": ["-C", "/path/to/reachy-mini-mcp", "run", "python", "src/server.py"],
      "env": {
        "DEEPGRAM_API_KEY": "your_key_here"
      }
    }
  }
}

Claude Code

~/.claude.json:

{
  "mcpServers": {
    "reachy-mini": {
      "command": "poetry",
      "args": ["-C", "/path/to/reachy-mini-mcp", "run", "python", "src/server.py"],
      "env": {
        "DEEPGRAM_API_KEY": "your_key_here"
      }
    }
  }
}

Note: The env block is required for speak() TTS. Get a free API key at deepgram.com.

Requirements

• Python 3.10+
• reachy-mini SDK (installed via poetry)
• MuJoCo (for simulation)
• Deepgram API key (for TTS, optional)

Hardware Notes

• Simulator: mjpython required on macOS for MuJoCo visualization
• Real hardware: Same MCP server, daemon auto-connects
• Port conflicts: Zenoh uses 7447, daemon uses 8765 by default

License

MIT License - see LICENSE

Acknowledgments

This project uses the Reachy Mini SDK by Pollen Robotics, licensed under Apache 2.0.

Links

• Reachy Mini SDK (Apache 2.0)
• FastMCP
• MCP Protocol
• mVara