OpenClaw model routing

Give OpenClaw one address for all your models

Connect OpenClaw to TEKIZ.AI locally, over your Tailnet, or through the hosted API. OpenClaw can then use Codex, Ollama, Gemini, NVIDIA NIM, and other accounts you connect without a separate setup for every agent.

Create API key OpenClaw guide Coding-agent routes Self-hosted routing

OpenClaw skill setup

Merge the TEKIZ.AI provider into OpenClaw’s native model configuration. Use the hosted endpoint with an active TEKIZ.AI key, or replace both endpoint and credential boundary with your local router.

Works with both common client formats

OpenClaw tools can use OPENAI_BASE_URL=http://localhost:8790/v1 or ANTHROPIC_BASE_URL=http://localhost:8790. Provider credentials stay on the router computer.

Use the Codex account already connected to OpenClaw

TEKIZ.AI can read OpenClaw-managed Codex authentication from the profile store or environment. It does not ask you to sign in through a separate device-login page.

Configure OpenClaw

Start the router in one terminal and leave it running. Use the client environment commands from a second terminal.

# Keep the key in a shell, secret manager, or service environment.
export TEKIZAI_API_KEY=sk_tekiz_your_key

# Validate the merged JSON before restarting a managed OpenClaw service.
openclaw config validate

# For generic OpenAI-compatible tools running beside OpenClaw:
export OPENAI_BASE_URL=https://api.tekiz.ai/v1
export OPENAI_API_KEY="$TEKIZAI_API_KEY"

Optional native provider object

Merge this object under models.providers in ~/.openclaw/openclaw.json. It is JSON only. Do not paste it into a terminal or overwrite your existing configuration.

{
  "tekizai": {
    "baseUrl": "https://api.tekiz.ai/v1",
    "apiKey": "${TEKIZAI_API_KEY}",
    "api": "openai-completions",
    "models": [
      { "id": "tekizai/auto", "name": "TEKIZ.AI Auto" },
      { "id": "tekizai/frontier", "name": "TEKIZ.AI Frontier" },
      { "id": "tekizai/fusion", "name": "TEKIZ.AI Fusion" }
    ]
  }
}

Prefer local models when they fit

Keep Ollama and private Tailnet models first, then use a connected cloud provider when the task needs more context or a stronger model.

Credential load balancing across your keys

Add more than one key or account. TEKIZ.AI shares requests across those that are working and pauses only the one that fails.

Use accounts you are allowed to use

TEKIZ.AI does not grant access to OpenAI, Codex, Anthropic, Gemini, NVIDIA, Ollama Cloud, or other providers. Connect your own account, or use a hosted TEKIZ.AI plan.

Keep model setup in one place

OpenClaw can use local models, cloud APIs, coding subscriptions, and file-capable models. TEKIZ.AI keeps that setup behind one address and chooses an available backup when a model or key fails.

Next steps

Create an API key, open the OpenClaw guide, copy the hosted quickstart, or estimate routing fit.