How to Connect Ollama on a Remote Machine to OpenClaw

A question that comes up regularly in the Discord: "How do I connect Ollama running on one PC to OpenClaw running on another?"

This is a common setup when you have a powerful GPU machine for inference but want to run OpenClaw on a lighter device (like a Mac Mini or VPS). Here's the complete guide.

The Problem

By default, Ollama only listens on localhost:11434. If OpenClaw is on a different machine, it can't reach Ollama at all.

Step 1: Configure Ollama to Accept Remote Connections

On the machine running Ollama, you need to set the OLLAMA_HOST environment variable to listen on all interfaces:

Linux (systemd)

# Edit the Ollama service
sudo systemctl edit ollama.service

# Add these lines:
[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"

# Restart Ollama
sudo systemctl daemon-reload
sudo systemctl restart ollama

macOS

# Stop Ollama if running as an app
# Then start from terminal with:
OLLAMA_HOST=0.0.0.0:11434 ollama serve

Windows

Set the environment variable OLLAMA_HOST to 0.0.0.0:11434 in System Properties → Environment Variables, then restart Ollama.

Step 2: Verify Ollama is Accessible

From your OpenClaw machine, test connectivity:

curl http://<OLLAMA_IP>:11434/api/tags

You should see a JSON list of your installed models.

Step 3: Configure OpenClaw to Use Remote Ollama

In your openclaw.json, set the baseUrl for the Ollama provider:

{
  "models": {
    "providers": {
      "ollama": {
        "baseUrl": "http://<OLLAMA_IP>:11434"
      }
    }
  }
}

Replace <OLLAMA_IP> with the actual IP address of your Ollama machine (e.g., 192.168.1.100).

Step 4: Set Your Model

Now configure OpenClaw to use an Ollama model:

openclaw models default ollama/qwen2.5:32b

Or via config:

{
  "agents": {
    "defaults": {
      "model": "ollama/qwen2.5:32b"
    }
  }
}

Security Considerations

Warning: Exposing Ollama to 0.0.0.0 makes it accessible to anyone on your network. For production setups:

Use a firewall to restrict access to only your OpenClaw machine's IP
Use a VPN like Tailscale or WireGuard for secure cross-network access

Consider SSH tunneling for temporary secure connections:

ssh -L 11434:localhost:11434 user@ollama-machine
# Then use localhost:11434 in OpenClaw

Troubleshooting

Connection refused: Check firewall rules on the Ollama machine
Timeout errors: Ensure both machines are on the same network or properly routed
Model not found: Verify the model is installed on the Ollama machine with ollama list

Bonus: Multiple Ollama Servers

You can even route different models to different machines by setting up multiple providers:

{
  "models": {
    "providers": {
      "ollama-gpu": {
        "type": "ollama",
        "baseUrl": "http://192.168.1.100:11434"
      },
      "ollama-cpu": {
        "type": "ollama",
        "baseUrl": "http://192.168.1.101:11434"
      }
    }
  }
}

Originally asked by Mr. Dante in the OpenClaw Discord. Got questions? Head to #users-helping-users!