
Managed Agents in the Gemini API can now run their tools in your own sandbox. While the agent loop and the model stay on Google's platform, shell and file tools run in a worker you operate. All connected via the Gemini API CLI over outbound HTTPS. For production I'd pick one fresh sandbox per interaction that holds a short-lived lease token and gets deleted when the session ends. Launch partners include Vercel, Cloudflare, Modal, Daytona, and of course, Google Cloud.
building with Managed Agents on the Gemini API? now you can bring your own sandbox with self-hosted environments start building today: ai.studio














