Setup local agent on Raspberry Pi 5 with OpenClaw and Ollama

Install Ollama

1. On the machine which will host the Ollama, create a folder for ollama.

   mkdir ollama
   cd ollama

2. Create docker-compose.yaml in ollama folder with content:

   services:
     ollama:
       image: ollama/ollama:latest
       container_name: ollama
       ports:
         - "11434:11434"
       volumes:
         - ./data:/root/.ollama
       restart: unless-stopped

3. Pull a model for local use, for example, qwen3:0.6b:

   docker compose exec ollama ollama pull qwen3:0.6b

Install OpenClaw

1. On Raspberry Pi, get OpenClaw source code from: https://github.com/openclaw/openclaw

2. Create docker:

   cd openclaw
   export OPENCLAW_SANDBOX=1
   export OPENCLAW_IMAGE=ghcr.io/openclaw/openclaw:latest
   ./docker-setup.sh

3. Open config file ~/.openclaw/openclaw.json, in section gateway, add controlUi field:

   {
     "gateway": {
       "controlUi": {
         "allowedOrigins": [
           "http://localhost:18789",
           "http://127.0.0.1:18789",
           "http://<your_pi_address>:18789"
         ]
       }
     }
   }

4. Restart gateway:

   docker compose restart openclaw-gateway

Add local Ollama to OpenClaw

Open ~/.openclaw/openclaw.json

1. In section models > providers, add information for the local Ollama:

   {
     "models": {
       "mode": "merge",
       "providers": {
         "ollama": {
           "baseUrl": "http://<ollama_machine_address>:11434",
           "api": "ollama",
           "models": [
             {
               "id": "qwen3:0.6b",
               "name": "qwen3:0.6b",
               "reasoning": false,
               "input": [
                 "text"
               ],
               "cost": {
                 "input": 0,
                 "output": 0,
                 "cacheRead": 0,
                 "cacheWrite": 0
               },
               "contextWindow": 16000,
               "maxTokens": 512,
               "compat": {
                 "thinkingFormat": "qwen"
               }
             }
           ]
         }
       }
     }
   }

2. Restart gateway:

   docker compose restart openclaw-gateway

3. Now your agent can use ollama/qwen3:0.6b as its model.