Expanding Managed Agents in Gemini API: background tasks, remote MCP and more
Google expands Gemini API's Managed Agents with background execution, remote MCP integration, custom functions, and credential refresh for improved developer workflows.
Google is enhancing the Gemini API by introducing new capabilities for Managed Agents, including background execution for asynchronous interactions, direct access to remote MCP servers, and support for custom functions and credential refresh. These updates aim to improve developer efficiency and reliability in building production-ready agents.
The background execution feature allows developers to run long-running tasks asynchronously, with the API returning an ID for status tracking. This enables clients to poll for progress or reconnect later, ensuring smoother handling of extended operations without blocking the application.
Managed agents can now connect directly to remote MCP servers, allowing them to interact with private databases or internal APIs without custom middleware. Developers can combine remote tools with built-in sandbox capabilities, enhancing flexibility and security in agent interactions.
Credential management has been simplified with the ability to refresh access tokens or rotate API keys by providing an existing environment_id and new network configuration. These updates ensure that managed agents operate in real development environments, maintaining state and installed packages while improving asynchronous workflows.