About the service
Google's agent-first development environment: you describe the task and the agent plans the steps, edits files, runs commands and verifies the result in a browser. You work through an editor, an agent manager and a separate CLI, so several tasks can run in parallel and changes are accepted after reviewing the agent's reports. Gemini is the core model, with Claude and GPT-OSS also available in the model selector under their own limit. Access is not bought on its own: it comes with a Google AI subscription, and the plan only determines quota size and priority. Usage is metered by how much work the agent does rather than by message count, so straightforward tasks drain the limit far more slowly.