Google adds background execution, remote MCP integration, and credential refresh to Managed Agents in Gemini API

Google announced new capabilities for Managed Agents in the Gemini API, including background execution for asynchronous interactions, remote Model Context Protocol (MCP) server integration, custom function calling, and credential refresh across interactions.

Background execution allows developers to run interactions asynchronously on the server by passing a background flag, with the API returning an ID for polling status, streaming progress, or reconnecting later while the agent completes tasks remotely.

Remote MCP server integration enables managed agents to connect directly to remote MCP servers, allowing agents to interact with private databases or internal APIs without writing custom proxy middleware.

Custom function calling lets developers add tools alongside built-in sandbox tools, with the system using step matching to determine whether built-in tools run on the server or custom functions trigger local execution via requires_action.

Network credential refresh supports refreshing expired access tokens or short-lived API keys by passing an environment ID with a new network configuration, replacing old rules immediately while preserving sandbox filesystem state, installed packages, and cloned repositories.