madisonjcoleman/clawmanager repository overview

⁠ClawManager

A web dashboard for deploying, managing, and monitoring OpenClaw⁠ AI agents on remote VMs over SSH.

⁠What it does

• Manage multiple OpenClaw nodes from a single dashboard
• Deploy OpenClaw to remote Linux VMs via SSH
• Built-in web terminal for SSH access to each node
• Edit openclaw.json and prompt files remotely
• UFW firewall management with preset rule templates
• Caddy reverse proxy deployment
• Real-time node health monitoring over WebSocket
• Traffic monitoring for agent HTTP/HTTPS requests
• Encrypted credential storage (AES-256-GCM)

⁠Running with Docker Compose

Clone the repo and run:

docker compose up -d

Or pull directly:

docker pull madisonjcoleman/clawmanager:latest

Then use docker-compose.yml from this repo, or write your own. The dashboard runs on port 12445 by default.

⁠Requirements

• Expose port 12445 (or set CLAWMANAGER_PORT)
• Mount a volume to /root/.openclaw for persistent data (nodes, credentials, configs)

⁠Environment Variables

⁠Persistence

Mount a host directory to /root/.openclaw to keep your data across restarts:

-v ~/.openclaw:/root/.openclaw

This stores the SQLite database (node definitions, encrypted SSH credentials), secrets, and configuration files.

⁠Networking

Standard mode (docker-compose.yml) uses bridge networking with host.docker.internal mapped, so the container can reach services on the host machine.

Host network mode (docker-compose.host.yml) runs directly on the host interface for full LAN access, useful when ClawManager needs to reach local inference servers or other LAN resources.

⁠Architecture

• Backend: Node.js, Express, Socket.io, ssh2, better-sqlite3
• Frontend: React, Vite, TailwindCSS, xterm.js
• Single container, multi-stage Docker build

License: MIT