Back to AI Q&A
What should I do if OpenClaw prompts Model not allowed?

What should I do if OpenClaw prompts Model not allowed?

AI Q&A Admin 148 views

OpenClaw (Lobster) hints Model "provider/model" is not allowed, meaning that the model you choose is not on the agents.defaults.models whitelist. It intercepts before generating a reply, so it looks like "lobster didn't reply to message". The workaround is not to reinstall, but to fix the model allowlist or switch back to the allowed model.

First, check the currently available models

  1. Use /model to view the available models in the chat.
  2. Command line execution openclaw models status see provider and authentication status.
  3. Check that the agents.defaults.model.primary and agents.defaults.models in the configuration are consistent.

Two amendments

  • Want to continue with this model: add it to the agents.defaults.models and confirm that the provider auth is available.
  • Just cut the model by mistake: use /model to cut back to the existing model in the list.

Don't delete the allowlist to save trouble. The value of whitelisting is to prevent sessions from being temporarily cut to models that are unreliable, too expensive, or don't support tool calls. Especially when open group chats or multi-person use, the clearer the range of model availability, the less likely it is to have cost and security issues.

Official open source address: https://github.com/openclaw/openclaw.

If it is a team environment, you should also write the optional model into the O&M documentation. Otherwise, if one person temporarily adds a new model and another fails to switch in the session, they will mistakenly believe that something is wrong with the Gateway, channel, or model provider.

Recommended Tools

More