Use Google Gemini with OpenClaw

OpenClaw supports Google models through the active Google provider integration. Use the provider’s supported AI Studio API-key path unless your deployment documentation calls for a different Google authentication route.

Configure the credential

Open the Control UI Model Providers page or run the current onboarding/configuration flow:

openclaw models status
openclaw models list
openclaw configure

Enter the key through the provider flow instead of committing it to openclaw.json. If the Gateway is managed as a service, a key exported only in your interactive shell will not automatically reach that service.

Select an exact model

List the Google models visible to the installed provider and keep the full provider/model reference. Model availability and names can change, so do not copy an old unqualified display name from a blog post.

Use the Control UI connection test before making the model primary. Then inspect the resolved route:

openclaw models status
openclaw config get agents.defaults.model

Verify tool behavior

Run one text-only turn, one harmless tool call, and one invalid-parameter recovery task. A model that answers normal chat is not automatically suitable for browser or filesystem automation.

Expected result

• The Google provider connection test succeeds.
• The selected model resolves with its provider prefix.
• A fresh session uses the intended route.
• Required tools produce valid structured arguments.

If the provider test returns an authentication or quota category, fix that result before changing models. Use Model failover only after the Google route works independently.