Google

Google provides the Gemini model family. AgentXchain connects via the api_proxy adapter using Google's Generative AI API.

Which adapter?

api_proxy with provider: "google" — AgentXchain sends governed turn prompts to Google's generateContent endpoint.

Prerequisites

• A Google AI API key — get one from Google AI Studio
• GOOGLE_API_KEY set in your environment
• agentxchain CLI installed

Configuration

{
  "runtimes": {
    "gemini-dev": {
      "type": "api_proxy",
      "provider": "google",
      "model": "gemini-3.1-pro",
      "auth_env": "GOOGLE_API_KEY"
    }
  },
  "roles": {
    "dev": {
      "runtime": "gemini-dev",
      "mandate": "Implement features and fix bugs",
      "authority": "proposed"
    }
  }
}

Minimal working example

agentxchain init --governed --template api-service --goal "Build a feedback intake API" --dir my-project -y
cd my-project
# Replace the scaffolded runtime wiring in agentxchain.json with the Google config above.
agentxchain doctor
agentxchain connector check
agentxchain connector validate gemini-dev
agentxchain run

If you prefer the guided interactive scaffold, run agentxchain init --governed without -y, then update agentxchain.json with the Google config above before agentxchain connector check and agentxchain connector validate gemini-dev.

Available models

Model	Best for
gemini-3.1-pro	Most capable — complex coding, long context
gemini-3.0-flash	Fast balanced model
gemini-3.1-flash-lite	High-volume, cost-sensitive roles

Gemma (open-weight)

Google's open-weight Gemma models can be run locally via Ollama:

ollama pull gemma4

{
  "runtimes": {
    "gemma-dev": {
      "type": "api_proxy",
      "provider": "ollama",
      "model": "gemma4",
      "auth_env": "OLLAMA_API_KEY"
    }
  }
}

Vertex AI (enterprise)

For Google Cloud enterprise deployments using Vertex AI:

{
  "runtimes": {
    "vertex-dev": {
      "type": "api_proxy",
      "provider": "google",
      "model": "gemini-3.1-pro",
      "auth_env": "GOOGLE_API_KEY",
      "base_url": "https://us-central1-aiplatform.googleapis.com/v1/projects/YOUR_PROJECT/locations/us-central1/publishers/google/models/gemini-3.1-pro:generateContent"
    }
  }
}

Verify the connection

export GOOGLE_API_KEY="AIza..."
agentxchain connector check
agentxchain connector validate gemini-dev

Gotchas

• API key vs. OAuth: The auth_env approach uses API keys (simplest). For Vertex AI with service accounts, you may need a wrapper that exchanges OAuth tokens.
• Long context: Gemini Pro supports very large context windows (1M+ tokens), making it well-suited for large dispatch bundles with extensive codebase context.
• Safety filters: Google's API applies safety filters that may block certain prompts. If a governed turn fails with a safety filter block, the adapter reports it as an extraction failure with the block reason.