Connect OauthRouter to OpenClaw as a custom OpenAI-compatible provider. Takes under 1 minute.
Go to oauthrouter.com/dashboard/api
In Your API Token, click Reveal and then Copy. The token starts with lr_live_.
In OpenClaw, navigate to Settings → Providers → Add Custom Provider.
Choose the provider type:
Both options work the same way.
| Field | Value |
|---|---|
| Name | OauthRouter |
| Base URL | https://api.oauthrouter.com/v1 |
| API Key | lr_live_YOUR_TOKEN |
| Default Model | anthropic/claude-sonnet-4-5 |
Click Save.
Create a new conversation in OpenClaw and select OauthRouter as the provider.
Pick a model:
anthropic/claude-sonnet-4-5 — balanced (recommended)anthropic/claude-haiku-4-5 — fast and cheapanthropic/claude-opus-4-6 — smartest (but slow, 60+ seconds)Send any message. If the model responds, you're connected.
Claude Opus 4.6 can take 60+ seconds for long responses. If using Opus, increase the OpenClaw client timeout:
For Sonnet and Haiku, the default 30 second timeout is fine.
The models you can use depend on which providers you configured in Dashboard → Providers. For example, if you only added Anthropic, only Claude models will work.
Check your available models dynamically:
curl https://api.oauthrouter.com/v1/models \
-H "Authorization: Bearer lr_live_YOUR_TOKEN"| Error | Cause | Solution |
|---|---|---|
| 401 authentication_error | Token invalid or expired | Go to /dashboard/api and click Regenerate token |
| 400 No enabled provider found | Using a model from a provider you have not configured | Add the provider (e.g. OpenAI) in /dashboard |
| 429 Monthly message limit reached | Reached your plan quota | Upgrade plan or wait until month reset (1st) |
| timeout | Slow model + short client timeout | Set OpenClaw timeout to 120s+ |
| Connection refused / DNS error | Wrong Base URL | Must be https://api.oauthrouter.com/v1 (with api. subdomain) |
| All models failed: 522 | Cloudflare reached timeout limit | Try again in a few seconds or use Haiku instead of Opus |
anthropic/claude-haiku-4-5FastestBest for quick responses, simple tasks, cheap. ~1-3 seconds per response.
anthropic/claude-sonnet-4-5RecommendedBest balance of speed and intelligence for most tasks. ~3-10 seconds.
anthropic/claude-opus-4-6SmartestBest for complex reasoning and coding. Slow — 30-90 seconds. Requires 120s+ timeout.