Google has updated its Gemini API to include Managed Agents, a feature that lets developers build and run autonomous AI agents. The expansion adds background execution, integration with the Model Context Protocol (MCP), and custom function calling.
What Managed Agents bring to the API
The new Managed Agents are designed to handle tasks without constant user input. Background execution means an agent can keep working even after a user closes the app or browser. That's a shift from the typical request-response pattern of most large language model APIs.
Google also added support for the Model Context Protocol, an open standard that lets agents pull in data from external tools and services. Developers can now define their own functions for the agent to call, giving them more control over how the agent interacts with other software.
The update is aimed at developers who want to build more complex, real-world automation. Instead of chaining together multiple API calls manually, a single Managed Agent can orchestrate a sequence of steps — fetching data, processing it, and triggering actions — all on its own.
Google says the enhanced API is intended to foster autonomous AI development. The company hasn't released specific pricing or availability details beyond what's in the API documentation.
What's still unclear
Google hasn't said how the Managed Agents handle errors or unexpected inputs during background execution. Developers will need to test those edge cases themselves. The company also hasn't disclosed any limits on how long an agent can run in the background or how many function calls it can make per session.
For now, the feature is available through the Gemini API. Developers can start experimenting with the new capabilities immediately.
