Google Gemini with OpenClaw

Google's Gemini models are supported in OpenClaw. Gemini 3.1 Pro (google/gemini-3.1-pro-preview) was added in OpenClaw 2026.2.21. Gemini Flash is cost-effective for heartbeats, cron jobs, and background tasks—many users run tiered routing with Claude or GPT for main chat and Gemini Flash for scheduled checks.

• Gemini 3.1 Pro — google/gemini-3.1-pro-preview (from 2026.2.21)
• Gemini Flash — Fast, inexpensive, good for heartbeats and background work
• Other Gemini variants — Check OpenClaw and Google AI docs for latest model IDs

Use Gemini Flash for tiered routing: reserve a stronger model for main chat; use Flash for "wake up and check" automation. See Automation for cron and heartbeats, and Example setups & model routing.

Configure an API key from Google AI Studio (aistudio.google.com) or the Vertex AI console. Store in ~/.openclaw/credentials or via openclaw secrets. OAuth/subscription is not typically used for Gemini with OpenClaw.

{
  "agent": {
    "model": "google/gemini-3.1-pro-preview"
  }
}

Or use google/gemini-flash for cost-effective background tasks. See Configuration guide for full options.

• Cost-effective — Flash is cheap for heartbeats and cron
• Tiered routing — Pair with Claude/GPT for main chat
• Google ecosystem — Integrates with Google AI Studio
• Fast iteration — New models added regularly

Gemini Flash works well for automation; enable Google Search grounding only when you need fresh web facts—it adds latency and cost. Pair with cron for scheduled checks.

Google AI keys leak via screenshots and shared repos. Rotate in Google Cloud console, update via openclaw onboard, restart Gateway, then revoke the old key.