shmayro/nemoclaw repository overview

⁠NemoClaw Docker

Run NVIDIA NemoClaw⁠ with a single command. NemoClaw is an open-source AI agent framework that runs OpenClaw securely inside NVIDIA OpenShell with managed inference.

⁠Quick Start

docker run -d --network host \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v nemoclaw-data:/nemoclaw-data \
  --name nemoclaw shmayro/nemoclaw

Open http://localhost:7681⁠ in your browser, then run:

nemoclaw onboard

⁠What's Inside

⁠Environment Variables

*At least one API key is needed to use NemoClaw. You can provide it at runtime or configure it during onboarding.

⁠Inference Providers

⁠Getting a Free NVIDIA API Key

1. Go to build.nvidia.com⁠
2. Sign in or create an account
3. Generate an API key
4. Pass it to the container: -e NVIDIA_API_KEY=nvapi-xxx

⁠Usage Examples

⁠With NVIDIA API (free, recommended)

docker run -d --network host \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -e NVIDIA_API_KEY=nvapi-xxx \
  -v nemoclaw-data:/nemoclaw-data \
  --name nemoclaw shmayro/nemoclaw

⁠With OpenRouter (Claude, GPT, Gemini, free models)

docker run -d --network host \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -e OPENROUTER_API_KEY=sk-or-xxx \
  -v nemoclaw-data:/nemoclaw-data \
  --name nemoclaw shmayro/nemoclaw

⁠With Web Terminal Authentication

docker run -d --network host \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -e NVIDIA_API_KEY=nvapi-xxx \
  -e TTYD_USER=admin -e TTYD_PASS=mysecret \
  -v nemoclaw-data:/nemoclaw-data \
  --name nemoclaw shmayro/nemoclaw

⁠Using Docker Compose

git clone https://github.com/shmayro/nemoclaw.git
cd nemoclaw
echo "NVIDIA_API_KEY=nvapi-xxx" > .env
docker compose up -d

⁠Access Methods

⁠Persistence

The /nemoclaw-data volume stores:

• NemoClaw configuration and API keys
• Onboarding state
• Sandbox data

Your data survives container restarts and upgrades.

⁠Security

• Host networking — The container uses --network host so NemoClaw's OpenShell gateway can communicate over localhost. This means ttyd is accessible on port 7681 of your host.
• Docker socket is mounted — The container uses the host's Docker daemon via /var/run/docker.sock. This gives the container access to manage Docker on the host. Only run on trusted machines.
• Web terminal is unauthenticated by default — Set TTYD_USER and TTYD_PASS to enable basic auth.
• Do not expose port 7681 to the internet without authentication.

⁠Requirements

⁠Telegram Bot

Connect NemoClaw to Telegram so you can chat with your AI agent from your phone.

⁠Setup

1. Open Telegram and message @BotFather⁠ with /newbot
2. Follow the prompts to create your bot and get a token
3. Add the token to your .env file or pass it as an env var:

docker run -d --network host \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -e NVIDIA_API_KEY=nvapi-xxx \
  -e TELEGRAM_BOT_TOKEN=your-bot-token \
  -v nemoclaw-data:/nemoclaw-data \
  --name nemoclaw shmayro/nemoclaw

1. Connect to the web terminal at http://localhost:7681 and run:

nemoclaw onboard                    # First time only
export TELEGRAM_BOT_TOKEN=your-bot-token
export NVIDIA_API_KEY=nvapi-xxx
export SANDBOX_NAME=my-assistant
nemoclaw start

1. Message your bot on Telegram — it responds via the NemoClaw agent!

⁠Telegram Environment Variables

⁠Stop the Bridge

nemoclaw stop

⁠NemoClaw Commands

Once inside the container:

nemoclaw onboard                    # Initial setup wizard
nemoclaw my-assistant connect       # Open sandbox shell
nemoclaw start                      # Start Telegram bridge
nemoclaw stop                       # Stop services
nemoclaw status                     # Check service status
openclaw tui                        # Interactive chat
openclaw agent --agent main -m "hello"  # CLI message
openshell term                      # Monitoring TUI

⁠Building from Source

git clone https://github.com/shmayro/nemoclaw.git
cd nemoclaw
docker build -t shmayro/nemoclaw .

⁠Contributing

Contributions welcome! Please open an issue or PR.

⁠License

MIT