Connect a local Ollama model
Your assistant runs in our data centre, the model runs on your machine, and Tailscale connects the two privately. Nothing is exposed to the internet.
You need
- A computer with a capable GPU or Apple Silicon that can stay on. OpenClaw needs a model of roughly 32B parameters or more to use tools reliably; small 7–8B models struggle.
- A free Tailscale account.
- A ClawNest account with a plan or an active trial.
On your computer
Install Ollama and pull a model
Install Ollama, then download a model:
Download Ollamaollama pull qwen3:32bLet Ollama listen on the network
By default Ollama only answers on localhost. Start it so it listens on all interfaces. On macOS and Windows you can set the same variable in the system environment and restart the app.
Ollama server configurationOLLAMA_HOST=0.0.0.0 ollama serveInstall Tailscale and note the IP
Install Tailscale on the same computer and log in. In the Tailscale admin console under Machines, note the computer’s IP address; it starts with
Tailscale machines100..
In ClawNest
Create an assistant with a local model
Open the full creation form at dash.clawnest.ai/bots/create?form=1. Give the assistant a name and choose Run a local model instead.
Open the creation formEnter the server and model
In Server URL (Tailscale IP), enter your computer’s Tailscale address with the Ollama port. In Model, type the exact model name you pulled, for example
qwen3:32b. Press Create Assistant.http://100.64.1.5:11434Join the assistant to your tailnet
When the assistant is ready, find the Local Network card on its Overview and choose Connect to Tailscale. Press Open Tailscale Console, generate an auth key under Settings → Keys, paste it into Tailscale Auth Key and press Connect to Tailscale. ClawNest does not store the key.
Test the connection and chat
Under Show more details, press Test next to LLM Connectivity. When it can reach your model, open the Chat tab and say hello.