OpenClaw + Ollama Quick Commands

Install Ollama (macOS)

brew install ollama
ollama --version

Start Ollama with this project's existing model directory

export OLLAMA_MODELS=/Users/simonking/OpenClawSandbox/ollama_data/models

ollama serve

Download models

ollama pull qwen2.5:7b
ollama pull mistral:7b
ollama pull nomic-embed-text:latest

Verify downloaded models

ollama list

Tailscale (secure remote access to OpenClaw)

Use Tailscale Serve to expose your local OpenClaw gateway on your tailnet.

1) Make sure Tailscale is connected

tailscale up

2) Publish OpenClaw port 18789 on your tailnet

tailscale serve --bg 18789

3) Check current Serve config

tailscale serve status

Expected result should look like:

https://<your-device>.<your-tailnet>.ts.net (tailnet only)
|-- / proxy http://127.0.0.1:18789

4) Reset and recreate Serve config (if needed)

tailscale serve reset
tailscale serve --bg 18789
tailscale serve status

Optional: public internet access

If you want internet-accessible exposure (not just tailnet), use Funnel:

tailscale funnel 18789

Note: Keep OpenClaw bound to loopback for safety; Tailscale Serve handles secure access.

Update OpenClaw to a new version

1) Pull the latest image and recreate the container

docker compose pull openclaw
docker compose up -d --force-recreate openclaw

Full clean restart (optional)

docker compose down
docker compose pull openclaw
docker compose up -d

Clean up old unused images (optional)

docker image prune -f

---

Connecting a new browser (device pairing)

When opening the Control UI in a new browser, you may see a "Device pairing required" screen.

1) List pending devices

docker exec openclaw-gateway node dist/index.js devices list

2) Approve the device

Copy the device ID from the pairing screen and run:

docker exec openclaw-gateway node dist/index.js devices approve <device-id>

Example:

docker exec openclaw-gateway node dist/index.js devices approve f8a7757c-52d3-42f0-af77-5ad217133775

3) Reconnect

After approval, click Reconnect (or reload the page) in your browser.

---

Offline mode: use Ollama as primary

docker exec openclaw-gateway node dist/index.js config set agents.defaults.model.primary ollama/qwen2.5:7b

Back to Bedrock when online

docker exec openclaw-gateway node dist/index.js config set agents.defaults.model.primary amazon-bedrock/openai.gpt-oss-20b-1:0

tmux quick use (keep Ollama running in background)

1) Create or attach a session

tmux new -s ollama
# or, if it already exists:
tmux attach -t ollama

2) Split panes

Ctrl+B then %    # vertical split (left/right)
Ctrl+B then "    # horizontal split (top/bottom)

3) Detach / reattach

Ctrl+B then D    # detach and keep processes running
tmux attach -t ollama

4) Run Ollama in one pane

export OLLAMA_MODELS=/Users/simonking/OpenClawSandbox/ollama_data/models
ollama serve