Which integrations are available for this server?

Provides tools for interacting with NVIDIA Isaac Lab, enabling RL training, environment management, session control, policy evaluation, and job monitoring on remote GPU instances.

How do I use mcp-server-isaaclab?

1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mcp-server-isaaclab Connect to my Brev GPU instance and list all available RL environments" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

mcp-server-isaaclab

by chloepilonv

Overview Schema Related Servers Score Discussions

Python

Local

mcp-server-isaaclab

MCP server for NVIDIA Isaac Lab — RL training, environment management, and policy evaluation from Claude.

Runs locally on your Mac and communicates with Isaac Lab on a remote Brev GPU instance through an SSH tunnel. All heavy simulation stays on the GPU; Claude just sends commands.

Not Isaac Sim. This server controls Isaac Lab (RL environments, training pipelines, policy evaluation). For low-level Isaac Sim control (USD prims, scene authoring, Kit commands), see mcp-server-isaacsim.

Architecture

┌──────────┐    stdio    ┌──────────────┐   SSH tunnel   ┌─────────────────┐
│  Claude   │◄──────────►│  MCP Server  │◄──────────────►│  Remote Agent   │
│  (local)  │            │  (local Mac) │   port 8421    │  (Brev GPU)     │
└──────────┘            └──────────────┘                └────────┬────────┘
                                                                  │
                                                          ┌───────▼────────┐
                                                          │   Isaac Lab    │
                                                          │  (Isaac Sim)   │
                                                          └────────────────┘

MCP Server (this repo) runs on your Mac as a stdio MCP server. It opens an SSH tunnel to the Brev instance and forwards all requests to the Remote Agent — a FastAPI service running next to Isaac Lab on the GPU box.

Related MCP server: MCP Remote Access

Prerequisites

Python 3.10+
A Brev GPU instance (provision with brev create)
SSH access to the instance (brev ssh)
Isaac Lab installed on the instance (setup script included)

Quick Start

1. Install locally

git clone git@github.com:chloepilonv/mcp-server-isaaclab.git
cd mcp-server-isaaclab
pip install -e .

2. Provision a Brev GPU instance

brev create isaaclab-gpu --gpu A100
brev ssh isaaclab-gpu

3. Install Isaac Lab on the instance

# From your Mac:
scp scripts/setup-brev-isaaclab.sh ubuntu@<BREV_HOST>:~
ssh ubuntu@<BREV_HOST> bash ~/setup-brev-isaaclab.sh

This installs Isaac Lab + the skrl, rsl_rl, and sb3 RL frameworks.

4. Deploy the remote agent

./scripts/deploy-remote-agent.sh <BREV_HOST> ubuntu ~/.ssh/your_key

This copies the agent code, installs it, and starts it as a systemd service on port 8421.

5. Configure Claude

The project includes .mcp.json so Claude Code automatically picks up the server when you're in this directory.

For Claude Desktop, add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "isaaclab": {
      "command": "mcp-server-isaaclab"
    }
  }
}

For Claude Code in other projects, add to the project's .mcp.json:

{
  "mcpServers": {
    "isaaclab": {
      "command": "mcp-server-isaaclab"
    }
  }
}

Tools

Connection

Tool	Description
`connect_instance`	Establish SSH tunnel to Brev GPU instance
`disconnect_instance`	Tear down the connection
`instance_status`	GPU utilization, active sessions & jobs
`gpu_status`	Detailed GPU memory, temperature, utilization

Simulation (Interactive)

Tool	Description
`list_environments`	List all registered Isaac Lab tasks
`create_session`	Create an interactive simulation session
`step_session`	Step simulation forward (random or specified actions)
`reset_session`	Reset environment to initial state
`get_observation`	Get current observations + action/obs space info
`close_session`	Close session and free GPU memory

Training

Tool	Description
`start_training`	Launch an async RL training job
`monitor_training`	Get status, recent logs, latest checkpoint
`get_training_logs`	Read full training logs
`stop_training`	Stop a running training job
`list_training_jobs`	List all jobs (running, completed, failed)

Evaluation & Files

Tool	Description
`evaluate_policy`	Evaluate a checkpoint, optionally record video
`list_checkpoints`	Browse saved model checkpoints
`list_log_dirs`	Browse training log directories
`list_videos`	List recorded simulation videos
`read_remote_file`	Read any text/image file on the instance
`run_isaaclab_script`	Run arbitrary Isaac Lab Python scripts

Example Conversations

Train a locomotion policy:

> Connect to my Brev instance at 203.0.113.42
> What environments are available for quadruped locomotion?
> Train Anymal-D on rough terrain with rsl_rl, 4096 envs, 1500 iterations
> Check on the training
> Evaluate the best checkpoint and record a video

Explore an environment interactively:

> Connect to my Brev GPU
> Create a session with Isaac-Cartpole-v0, 32 envs
> What does the observation space look like?
> Step 100 times with random actions — what are the rewards?
> Reset and try again
> Close the session

Monitor GPU and manage jobs:

> What's the GPU status?
> List all training jobs
> Stop the Ant training — it's not converging
> Show me the last 200 lines of logs from the Franka training

Supported RL Frameworks

Framework	Best For	Notes
skrl	General purpose	Modern, modular, good default choice
rsl_rl	Locomotion	ETH RSL's framework, optimized for legged robots
sb3	Prototyping	Stable Baselines 3, easy to use
rl_games	Multi-GPU	NVIDIA's framework, scales well

Available Environments (selection)

Category	Examples
Classic	`Isaac-Cartpole-v0`, `Isaac-Ant-v0`, `Isaac-Humanoid-v0`
Manipulation	`Isaac-Reach-Franka-v0`, `Isaac-Lift-Cube-Franka-v0`, `Isaac-Open-Drawer-Franka-v0`
Locomotion	`Isaac-Velocity-Flat-Anymal-D-v0`, `Isaac-Velocity-Rough-Unitree-Go2-v0`
Navigation	`Isaac-Navigation-Flat-Anymal-C-v0`

Use list_environments to get the full list from your installation.

Project Structure

mcp-server-isaaclab/
├── src/mcp_server_isaaclab/
│   ├── server.py            # MCP server (runs locally, exposes tools)
│   ├── connection.py        # SSH tunnel + HTTP client manager
│   └── remote/
│       └── agent.py         # FastAPI agent (runs on Brev GPU)
├── scripts/
│   ├── deploy-remote-agent.sh      # Deploy agent to Brev
│   └── setup-brev-isaaclab.sh      # Install Isaac Lab on instance
├── .mcp.json                # Claude Code MCP config
├── pyproject.toml
└── README.md

Development

pip install -e ".[dev]"
ruff check src/
pytest

License

MIT

This server cannot be installed

license - permissive license

quality - not tested

maintenance

How are these scores calculated?

Maintenance

–Maintainers

–Response time

–Release cycle

–Releases (12mo)

Commit activity

Resources

GitHub Repository

Need Help?

Related Servers

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Latest Blog Posts

Your AI Chatbot Just Exposed Your CEO's Salary to an Intern
By Om-Shree-0709 on July 2, 2026.
Agent Identity
MCP Security
OAuth Delegation
Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)
By Om-Shree-0709 on June 30, 2026.
Agentic Ai
Prompt Injection
WebAssembly
Lightport: Open-Sourcing Glama's AI Gateway
By punkpeye on April 27, 2026.
OpenAI
open source

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/chloepilonv/mcp-server-isaaclab'

If you have feedback or need assistance with the MCP directory API, please join our Discord server