Google’s Gemini API gets background execution and MCP support
Google DeepMind is extending its Managed Agents in the Gemini API with four new capabilities that let agents run asynchronously, link directly to remote MCP servers, combine custom functions with sandbox tools, and refresh credentials without losing workflow state.
A step toward more resilient agents
Managed Agents can now execute tasks in the background, allowing developers to queue work that runs independently of the main thread. This means long-running jobs no longer block user interactions or require persistent foreground sessions. The update also introduces native support for connecting to remote Model Context Protocol (MCP) servers, giving agents direct access to external tools and data sources without intermediate orchestration layers.
Credentials that survive refresh cycles
Another new feature keeps agents functional even when their credentials are rotated. Previously, credential refreshes could disrupt running workflows, forcing developers to rebuild state or restart processes. With this change, Managed Agents can now maintain continuity, preserving context across credential updates—a practical improvement for production deployments where security policies mandate regular rotation.
What it means for developers and users
These additions lower the barrier for deploying robust, always-on agents in applications that demand reliability and external connectivity. Teams building workflow automation, customer support bots, or internal tool integrations can now rely on background execution to handle background tasks and MCP for seamless tool access, reducing the need for custom bridging code.
Why it matters
For teams already using or planning to adopt Managed Agents, background execution and MCP support translate into fewer integration headaches and more predictable uptime. The credential refresh resilience also aligns with enterprise security practices, making the API more viable for production environments where stability and compliance are non-negotiable. In short, these updates push the Gemini API closer to being a turnkey platform for agentic workflows rather than a set of discrete, stateful functions.
Source: The Decoder. AI-assisted editorial synthesis — TechnoExpress.