# lpigeon/ros-mcp-server [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/lpigeon/ros-mcp-server  
**GitHub Stars:** 1507  
**Views:** 1  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/lpigeon-ros-mcp-server

## Description
The ROS MCP Server supports robot control by converting user-issued natural language commands into ROS or ROS2 control commands.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What lpigeon/ros-mcp-server MCP server does

The lpigeon/ros-mcp-server MCP server exposes ROS and ROS 2 robot interfaces to an MCP-compatible language-model client. It is intended for systems where an agent needs to both send commands and inspect the robot's current state.

The server is designed to work with existing robot applications. Adding a rosbridge node to the ROS setup is described as sufficient to connect the MCP layer, so the robot's source code does not need to be changed. Compatibility covers ROS 1 distributions and ROS 2 distributions, including Jazzy and Humble.

## How it works

An MCP client sends natural-language requests to the server. The server exposes the robot's ROS interfaces so the model can identify available topics, services, actions, and their types before using them. The README specifically notes that this discovery includes custom interface types and the syntax needed to use them.

Communication is bidirectional. An agent can issue an operation such as publishing a command or calling a service, then observe resulting topics, sensor readings, or other state changes. The lpigeon/ros-mcp-server MCP server therefore fits both direct robot control and diagnostic workflows, provided the underlying ROS system exposes the required interfaces through rosbridge.

The repository describes examples involving ROS 2 navigation and manipulation, an industrial robot end effector, and a simulated Unitree Go2 in NVIDIA Isaac Sim. These examples demonstrate possible deployments rather than a fixed set of supported robot models.

## Setup and configuration

Installation is covered by the repository's installation guide. The ROS environment must include a rosbridge node that connects the robot's existing ROS graph to the MCP server. The project advertises Python 3.10 or newer and pip 23.0 or newer, and it is marked as dev-container ready.

After the ROS-side bridge is available, configure an MCP-compatible client to launch or connect to the server according to the installation documentation.

## Tools and capabilities

The lpigeon/ros-mcp-server MCP server provides capabilities for:

- Publishing messages to ROS topics.
- Subscribing to topics and observing incoming data.
- Calling ROS services and actions.
- Reading and changing ROS parameters.
- Inspecting sensor data and robot state.
- Discovering interface names, types, and custom message syntax.

Agents and developers should consult the repository's current documentation before building calls that depend on specific schemas.

## Limitations and notes

The server does not replace the robot's ROS application or define a universal robot command vocabulary. What an agent can control depends on the topics, services, actions, parameters, and message types available in the connected ROS graph. Safe operation also depends on the robot, its ROS configuration, and the permissions or safeguards outside the information provided here.

ChatGPT does not have a matching client slug in this directory's supported-client list, so it is not represented in the metadata below. The project is licensed under Apache License 2.0.

_Full upstream README: https://allmcps.com/mcp/lpigeon-ros-mcp-server/readme_