Skip to main content
OpenClaw is a self-hosted, open-source personal AI agent you reach from WhatsApp, Telegram, Discord, and other messaging platforms, with MCP tool support. It picks its model from a providers config — add a custom OpenAI-compatible provider pointed at https://api.tokios.com/v1 with your sk-tok-… key, and OpenClaw’s agent runs on a model on your own hardware.

​
Prerequisites

Connector paired

Install and pair tokios-connector next to the model you want OpenClaw to reach.

Model registered

Register a deployment (public name) for that model.

API key created

Create an sk-tok-… key, optionally scoped to that deployment.

Supported backends

Confirm your local backend (Ollama, llama.cpp, vLLM, LM Studio) is running.

​
Add Tokios as a model provider

1

Open your OpenClaw config

Edit OpenClaw’s configuration, where model providers are defined under models.providers. The exact file path depends on your install — check your OpenClaw setup.
2

Add a Tokios provider

Add a provider entry with api set to openai-completions, baseUrl pointed at Tokios, and your deployment listed under models:
The id under models is your public deployment name from the Models tab, not your backend’s local model id.
3

Make it the agent's model

Point the agent’s default model at the provider you just added, using the <provider>/<model-id> form:
4

Provide the API key

Set TOKIOS_API_KEY in your environment to your sk-tok-… key rather than pasting it into the config file, then restart OpenClaw so it reloads the provider.
Replace gemma-tunnel with the deployment name you registered, and set TOKIOS_API_KEY to your own sk-tok-… key. Never commit a real key into the config file. OpenClaw can also speak the Anthropic surface — set api: "anthropic-messages" with baseUrl: "https://api.tokios.com" (no /v1) if you prefer it.

​
Your assistant, your model

OpenClaw is the agent you message from your phone; Tokios gives whatever model you’re running a stable, authenticated endpoint. The connector next to that model dials out to Tokios, and OpenClaw calls https://api.tokios.com/v1 with your key — so your personal assistant runs on a model you host, without an inbound port or VPN.

​
Troubleshooting

The API key OpenClaw sent is missing, malformed, or has been revoked. Confirm TOKIOS_API_KEY holds the full sk-tok-… key with no extra whitespace, and check its status on the Keys tab.
The model id doesn’t match a registered deployment, or baseUrl is missing /v1. Confirm the models id matches the public name on the Models tab, and that baseUrl is https://api.tokios.com/v1.
The connector for that deployment is offline or at its concurrency cap. Check the Connectors tab — the connector needs to show as Online, and the machine running it needs to still be up and reachable.

​
Next steps

Compatible clients

See base URL and key setup for other OpenAI- and Anthropic-compatible clients.

API keys

Create, scope, and rotate sk-tok-… keys for teammates and agents.