OpenClaw AI Gateway — Deployment Guide

View source on GitHub
Preset

OpenClaw AI Compute Gateway

Deploy OpenClaw AI messaging gateway, with optional local AI model powered by your device's GPU.

DevicePurpose
reComputer J40Runs OpenClaw gateway and local AI model with GPU acceleration

What you'll get:

  • AI chatbot gateway supporting 20+ messaging platforms
  • Optional local AI model running on device GPU — no data leaves your network
  • Web management interface for configuration

Requirements: Docker installed · Internet access (for first-time image download)

1

Deploy OpenClaw

Deploy the OpenClaw AI gateway. If local AI model is enabled, it will be started and configured automatically.

After Deployment

OpenClaw AI gateway is deployed. Follow the "Deployment Complete" instructions in the step above to log in, then start using it.

Try a Conversation

  1. Click Chat in the left sidebar and send a message to verify AI is responding
  2. If local AI model is enabled, it's already configured — no extra setup needed

Connect Messaging Platforms

  1. Click Channels in the left sidebar to add your messaging platform (WeChat, Telegram, Discord, etc.)
  2. Follow the on-screen instructions to authorize the platform
  3. Send a test message through the connected platform

Next Steps

Deployment method

Deploy on your reComputer J40 device (running this software locally on the Jetson).

Wiring
  1. Ensure Docker is installed and running
  2. Optionally check Enable Local AI Model and select a model
  3. Click Deploy to start services
What you will be asked for(filled in the SenseCraft Solution app)
3 more with defaults you can keep
  • Local File Access (optional)default ~/openclaw_docs

    Let OpenClaw read files from this directory. Useful for letting AI reference local documents. Leave empty to skip.

  • Enable Ollama (Local LLM)default false

    Start Ollama alongside OpenClaw for local AI inference

  • Ollama Modeldefault qwen3.5:2b-q4_K_M
    3 options
    • · Qwen3.5 2B Q4 (1.9GB, fast)
    • · Qwen3.5 2B (2.7GB)
    • · Qwen3.5 4B (3.4GB)

    Model to pull and use with Ollama

Troubleshooting
IssueSolution
Port 18789 already in useStop the service occupying port 18789, or check if OpenClaw is already running
Docker not foundInstall Docker Desktop and ensure it is running
Model download slowLarge models take time; check your internet connection
OpenClaw container keeps restartingCheck logs: docker logs openclaw-gateway
After Deployment
  1. Open http://localhost:18789 in your browser (the gateway binds to loopback only — no token needed)
  2. Connect your first messaging channel (WeChat, Telegram, Discord, etc.)
  3. If local AI model is enabled, it's already configured — select it when creating an agent