Skip to main content
OpenWork works with Google’s Gemini models. For yourself, paste a Gemini API key under Settings > AI Providers > Connect provider and choose Google. For a team on Google Cloud, an OpenWork admin connects Vertex AI (now called Gemini Enterprise Agent Platform) once through OpenWork AI Gateway, and each member signs in with their own Google account.

​
Choose your path

​
Option 1: Gemini API key

  1. Create an API key in Google AI Studio. Copy it once and keep it private.
  2. In OpenWork, open Settings and choose AI Providers.
  3. Click Connect provider.
  4. Type google in Filter providers by name or ID and select Google.
  5. Paste the key into the API key field and save.
  6. Open the chat model picker and choose a Gemini model.
No OpenWork Cloud account is needed. Usage is billed by Google to the account that owns the key.

​
Option 2: Vertex AI for a team

Use this when your organization already pays for Google Cloud and wants members to use Gemini (and Claude on Google Cloud) without handing out API keys.
  1. An OpenWork organization admin follows Google Agent Platform (Vertex AI) to enable the API, grant IAM access, and create a Google OAuth app.
  2. In OpenWork AI Gateway, the admin adds the google-vertex provider for Gemini (or google-vertex-anthropic for Claude), enters the Google Cloud project ID and region, and picks which models members can use.
  3. Each member signs in with their own Google account from the provider’s connection page.
  4. Members pick the assigned models in the OpenWork model picker.
Requests are billed to your Google Cloud project, and Google IAM decides who can run inference. Self-hosted deployments first need to enable OpenWork Gateway.

​
Option 3: Your own gateway in front of Vertex AI

If your team already runs an OpenAI-compatible gateway (for example, LiteLLM) that forwards requests to Vertex AI, connect the gateway instead of Google directly:
  1. Get the gateway’s base URL, the model names it exposes for Gemini, and a key.
  2. For yourself, add it as a custom provider in the workspace opencode.jsonc. For the team, an admin adds it once as a custom LLM provider in OpenWork Cloud.
  3. If the gateway issues a separate key per person, use per-member LLM credentials.
See Add a custom LLM or AI gateway for the config format.

​
FAQ

OpenWork itself is free. Gemini usage is billed, or rate-limited, by Google according to the plan on your API key or Google Cloud project.
Yes. Google renamed Vertex AI to Gemini Enterprise Agent Platform. OpenWork still uses the provider IDs google-vertex and google-vertex-anthropic.
Not with Option 2. Each member signs in with Google through OpenWork, and Google IAM controls access.
Yes. Add google-vertex-anthropic in Option 2 after enabling the Claude model in Model Garden.
No. Model access and Google Workspace data are separate. To let the agent read Gmail, Calendar, or Drive, follow Connect your services.