Google Gemini API adds background tasks and remote MCP for managed agents

Google expanded Managed Agents in Gemini API this week with features designed to support production-scale agent deployments. The update adds background task execution, remote Model Context Protocol (MCP) support, and additional reliability tooling for developers building autonomous workflows. Background tasks allow agents to run long-duration operations asynchronously without blocking client connections—essential for data processing pipelines or scheduled monitoring jobs. Remote MCP support means agents can now connect to external context sources and tools over the network, extending the range of data and APIs an agent can access at runtime.

Managed Agents, introduced earlier this year, handle orchestration and state management server-side, shifting operational overhead from developers to Google's infrastructure. MCP, a standard protocol originally developed by Anthropic, defines how language models interact with external systems—databases, search engines, code execution environments—without custom integration code for each tool. The features are available now in the Gemini API.