ROS-MCP Server is now open source

We are excited to announce the public release of the ROS-MCP Server — an open-source tool that connects large language models like Claude, GPT, and Gemini with ROS and ROS2 robots through the Model Context Protocol.

What it does

The ROS-MCP Server gives LLMs full, bidirectional access to any ROS robot. With no changes to your existing robot code, you can:

• Command robots in natural language — instructions are translated into ROS topics, services, and actions, including custom message types
• Give AI full visibility — subscribe to topics, call services, read sensor data, and monitor robot state in real time

It works with both ROS1 and ROS2, on Linux, Windows, and macOS.

See it in action

We've published demos showing the server controlling real and simulated robots:

• Debugging an industrial robot — the LLM browses topics, queries custom service types, and calls services to test the robot
• Controlling a Unitree Go2 with natural language — Claude interprets camera images and commands the quadruped based on voice instructions

Get started

The server is available on GitHub under the Apache 2.0 license:

github.com/robotmcp/ros-mcp-server

We also announced the release on ROS Discourse — check it out and let us know what you think.

Contributions, bug reports, and new examples are all welcome.