Google Deepmind has expanded its Managed Agents in the Gemini API with four new features: asynchronous background execution that does not require an open HTTP connection, support for remote MCP servers to connect directly to databases and APIs, custom functions alongside sandbox tools, and automatic credential refresh. These capabilities are available immediately through the Gemini Interactions API.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Google Deepmind has added four new capabilities to Managed Agents in the Gemini API: Background Execution (allowing agents to run asynchronously without an open HTTP connection), direct connection to remote MCP servers for internal databases or APIs, support for custom functions alongside built-in sandbox tools, and automatic credential refresh between interactions.
Why it matters
These additions expand what developers can build with Gemini's agent framework, removing the constraint of keeping an HTTP connection open and enabling deeper integration with existing enterprise systems and data sources through MCP protocols.
What to watch
All features are available now through the Gemini Interactions API, with code examples provided for JavaScript, Python, and cURL in the documentation.
Google Deepmind's expansion of the Gemini API's Managed Agents addresses practical bottlenecks in agent deployment. The Background Execution feature removes a technical constraint—the requirement to maintain an open HTTP connection—that would have otherwise forced developers to keep resources allocated for the duration of an agent's work. This is particularly valuable for long-running tasks where maintaining an active connection would be inefficient.
The addition of remote MCP (Model Context Protocol) support signals a strategic move toward interoperability with existing enterprise systems. MCP is an emerging standard for connecting AI systems to external tools and data sources; by allowing agents to connect directly to internal databases and APIs through MCP, Google is positioning Gemini agents as components that can integrate into mature software stacks without heavy custom middleware. The ability to refresh credentials between interactions without resetting sandbox state preserves security and statefulness—two concerns that can otherwise force developers to choose between convenience and safety.
Together, these changes lower friction for developers building production agent applications, particularly those working within organizations with existing database and API infrastructure that they want agents to access safely and persistently.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
