Use Mission Control MCP Server in Cursor

Overview

This tutorial explains how to run the isaac_mission_control MCP server (mission_control_mcp) and connect it to Cursor. This allows you to control robot missions using natural language prompts in Cursor’s AI chat.

It uses direct MCP configuration in ~/.cursor/mcp.json and does not require a .env file.

Tutorial Walkthrough

1. Clone the isaac_mission_control repository and navigate into it:

   git clone https://github.com/nvidia-isaac/isaac_mission_control.git
   cd isaac_mission_control
   

   All subsequent commands in this tutorial (unless otherwise noted) are run from the root of this isaac_mission_control directory.

2. Complete the following prerequisites before continuing (steps a and b can be done in parallel):

   1. Isaac Sim: Install and launch Isaac Sim using the Isaac ROS Isaac Sim Setup Guide, then open the warehouse scene with Mission Client by following the Isaac ROS Mission Client Tutorial through launching Mission Client.

   2. Mission Control services: From the root of the isaac_mission_control repository, change into the compose directory and start the services:

      cd docs/tutorial/docker-compose
      docker compose -f bringup_services_tutorial.yaml up
      

      Note

      docker compose up runs in the foreground and holds this terminal open. Do not close this terminal. Open a new terminal tab before continuing with the next step.

3. In the new terminal tab, navigate to the isaac_mission_control root directory (the directory you cloned in step 1):

   cd /path/to/isaac_mission_control
   

   Then verify that isaac_mission_control is running:

   curl http://localhost:8050/api/v1/health
   

   Run all remaining steps in this tutorial from this new terminal, from the isaac_mission_control root directory.

4. Set up the MCP package. From the isaac_mission_control root:

   cd app/agentic-utilities/mission-control-mcp
   python3 -m venv venv
   ./venv/bin/python -m pip install -U pip
   ./venv/bin/python -m pip install -e .
   

5. Configure Cursor MCP directly in ~/.cursor/mcp.json (or use Cursor Settings -> Tools & MCP, which writes this file).

   

   

   Add or update this server entry with your real local paths:

   {
     "mcpServers": {
       "mission-control": {
         "command": "bash",
         "args": [
           "-lc",
           "set -euo pipefail; cd \"/ABS/PATH/TO/isaac_mission_control/app/agentic-utilities/mission-control-mcp\" && exec \"/ABS/PATH/TO/isaac_mission_control/app/agentic-utilities/mission-control-mcp/venv/bin/python\" -m mission_control_mcp.server"
         ],
         "env": {
           "MISSION_CONTROL_URL": "http://localhost:8050"
         }
       }
     }
   }
   

   Note

   Replace /ABS/PATH/TO/isaac_mission_control with your actual path. If isaac_mission_control is remote or port-forwarded, update MISSION_CONTROL_URL accordingly.

6. Enable the mission-control server in Cursor Settings -> Tools & MCP.

7. Start a new chat and ask:

   Use Mission Control MCP to test the connection.

   If configured correctly, the server responds with a successful connection result.

   

Isaac Mission Control 4.3.0 Limitations

The following isaac_mission_control MCP tools are not supported in Isaac Mission Control 4.3.0:

MCP Tool	Ready for Isaac Mission Control 4.3.0
submit_charging_mission	No
submit_undock_mission	No
get_detected_objects	No
get_detected_apriltags	No
visualize_route	No
get_map_info	No
list_available_maps	No
select_map	No
deploy_map_to_robot	No
submit_objective	No
cancel_objective	No

Troubleshooting

Connection Refused

• Confirm isaac_mission_control is reachable at the configured MISSION_CONTROL_URL.

Tools Do Not Appear

• Disable/re-enable the MCP server in Cursor settings.

• Start a new chat.

• Restart Cursor.

Check Server Logs

• /tmp/mission_control_mcp.log

Implementation References

• MCP server tools: app/agentic-utilities/mission-control-mcp/src/server.py

• isaac_mission_control API client used by MCP: app/agentic-utilities/mission-control-mcp/src/queries.py