Webots

You can connect Claude Code and Cursor to a Webots simulation through a dedicated MCP server that streams real-time robot state, sensors, camera frames, and control commands. This bridge lets your AI coach and tooling interact with the running simulation as if it were a live MCP endpoint, enabling tight feedback loops and automated control.

How to use

To use the Webots MCP Server, run the local MCP server alongside your Webots simulation and point your MCP clients to the provided stdio interface. The server exposes a set of state and control endpoints that your clients can publish to and read from, enabling real-time visibility into pose, sensors, camera data, and simulation controls. You can extend behavior with custom commands and hooks, such as reacting to world reloads or handling domain-specific actions.

How to install

Prerequisites: Python 3.10 or newer and pip for installing Python dependencies.

Option 1: Clone (Recommended) and install dependencies.

git clone https://github.com/luisfelipesena/webots-youbot-mcp.git
pip install mcp pydantic

Option 2: Copy files directly into your project folder.

# Copy mcp_bridge.py into your controller directory

Start the MCP server using the exact command and path shown in the quick start example, adjusting paths to your setup.

Configuration and usage tips

Configure your client to connect to the Webots MCP Server by specifying the Python command and the path to the server script, along with the working directory used during development or deployment.

Example usage snippet for including the server in your environment (as shown in code samples):

# In a client project

# Example: How to reference the MCP server from a client
# This reflects the provided integration approach

Advanced: World reload detection and custom commands

You can respond to Webots world reloads by registering a callback and triggering a reset of your client-side state. You can also register custom command handlers to execute project-specific actions when the MCP server receives a command from the bridge.

Global MCP configuration

Place MCP server settings in your central configuration so Claude Code and Cursor can locate and communicate with the server consistently.

{
  "mcpServers": {
    "webots": {
      "command": "python",
      "args": ["/path/to/webots-youbot-mcp/webots_youbot_mcp_server.py"],
      "cwd": "/path/to/webots-youbot-mcp"
    }
  }
}

Available tools

webots_get_robot_state

Get current robot state such as pose and mode from the simulation.

webots_get_sensors

Retrieve sensor data from the robot’s status.

webots_get_camera

Provide the latest camera frame path or data for processing.

webots_get_logs

Obtain controller logs for debugging and auditing.

webots_simulation_control

Control the simulation with actions like pause, resume, reset, reload, step, or speed changes.

webots_world_reload

Force a world reload to restart the simulation environment.

webots_world_reset

Reset the simulation to its initial state.

webots_reset_controller_state

Clear or reset the MCP state stored by the controller.

webots_take_screenshot

Capture and save a screenshot of the current simulation view.

webots_monitor

Monitor robot activity for a specified duration.

webots_get_full_state

Dump a complete state snapshot of the robot and environment.