Models and tools
Google Gemini
A Google Gemini integration stores a project-owned Google AI API key so Mobius-hosted agents can run Gemini models through that key.
Use Gemini when agents should use your Google AI account and quota. When media services are enabled, the same provider can also own built-in image and video actions.
Capability map
| Capability | Value |
|---|---|
| Provider ID | gemini |
| Auth kind | api_key |
| Connect flow | Inline API key |
| Actions | Deployment-dependent |
| Events | No |
| Webhook delivery | No |
| Event samples | No |
| Live status | No |
Connect Google Gemini
| Surface | Support |
|---|---|
| App | Open Govern > Integrations, choose Google Gemini, and paste a Google AI API key. |
| CLI | The mobius CLI does not connect Gemini integrations yet. Use the app or API. |
| API | Call POST /v1/projects/{project}/integrations/providers/gemini/connect with api_key. |
A project-owned key takes precedence over the platform credential for Gemini models. Media actions may also use Mobius-managed Google credentials when the catalog marks them ready through the platform source.
Actions
Gemini model credentials alone do not add action catalog entries. When media generation is enabled, these actions can appear:
gemini.edit_image
gemini.generate_image
gemini.generate_videoGenerated media is persisted as an artifact. Video generation is long-running, so expect the action step to stay active for several minutes.
Events
Google Gemini does not register source events.
Next
- Choose agent models from agents.
- Inspect generated media as artifacts.
- Track usage limits with guardrails.