This guide explains how OpenClaw handles model routing on ClawCloud — specifically the difference between managed mode (default) and bringing your own key.
What OpenRouter is
OpenRouter is a routing layer that sits in front of multiple AI providers. Instead of managing separate API keys for Anthropic, OpenAI, and Google, you make requests to a single OpenRouter endpoint and it routes them to the right model.
OpenClaw has native OpenRouter support. When you use managed mode on ClawCloud, your instance connects to OpenRouter — not directly to the model provider.
How managed mode works
When you deploy on ClawCloud without providing an API key, you're using managed mode. ClawCloud provisions a dedicated OpenRouter sub-key for your instance, with a credit limit that matches your plan:
- Lite — $8/month
- Pro — $25/month
- Max — $60/month
This sub-key is scoped to your instance only. Your bot uses that credit budget, and the counter resets at the start of each billing cycle. You can see current usage in the dashboard.
The sub-key is created automatically at checkout and pushed to your server during provisioning. You don't interact with OpenRouter directly — ClawCloud handles the key lifecycle.
Model naming in managed mode
OpenRouter uses slightly different model IDs than the native providers. OpenClaw translates these automatically, so the model you see in the interface is the model you get. There's no hidden routing or tier substitution.
If you run openclaw models list on your bot, the IDs shown are the OpenRouter-format IDs — for example, anthropic/claude-sonnet-4 rather than a provider-specific version string. The behavior and capabilities are identical to calling the provider directly.
BYOK — bringing your own key
If you already have an Anthropic, OpenAI, or Google API key, you can use it directly. This is the BYOK (bring your own key) mode.
With BYOK:
- Your key connects directly to the provider, not through OpenRouter
- Credit usage is billed to your provider account — ClawCloud doesn't track it
- No managed credit limit applies
- You control your own usage caps and spending alerts on the provider side
BYOK is set during the deploy wizard. You select the provider and paste your key. Switching modes after deployment requires re-provisioning.
Which mode to choose
Managed mode makes sense if you don't have an existing API account or want predictable monthly costs. The credit limits are conservative — they cover daily use on budget models, and moderate use on mid-tier models like Claude Haiku or GPT-4.1-mini.
BYOK makes sense if you already pay for API access at a provider, want to use models outside the managed catalog, or need no credit ceiling.
See the credits guide for a full breakdown of how credits are tracked and reset.
Troubleshooting
Bot says "model not found" after switching — In managed mode, use OpenRouter model IDs (e.g., anthropic/claude-sonnet-4). In BYOK mode, use native provider IDs (e.g., anthropic/claude-sonnet-4-5). See the model switching guide for the full alias list.
Credit usage seems higher than expected — OpenRouter charges per token processed, including system prompts sent with every message. Longer system prompts increase cost. See the credits guide for tips on reducing per-message cost.
Want to switch from managed to BYOK (or vice versa) — Mode is set during deployment and applies for the instance's lifetime. Switching requires re-provisioning. See Destroy vs Regenerate.
Deploy OpenClaw with Managed Credits