NewsBite analysis · AI, Programming, Tools
The Decoder
Google Deepmind adds background execution and MCP support to Gemini API managed agents
Google Deepmind has introduced significant enhancements to its Gemini API, aimed at improving the functionality and usability of Managed Agents. The new features include background execution, which allows agents to run asynchronously without a constant HTTP connection, enhancing application performance. Additionally, the integration of the Model Context Protocol (MCP) enables direct connections between MCP servers and internal databases or APIs, allowing for more efficient data management. Developers can also leverage custom functions alongside existing sandbox tools, and token management has been improved with capabilities to refresh credentials seamlessly while maintaining the sandbox state. This broadened functionality provides developers with greater flexibility and efficiency in building applications.
Act nowConfidence: HighShort term horizon
Published 8 Jul 2026, 16:10 UTC1 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- New background execution feature allows asynchronous agent processing
- MCP support enables direct database connections
- Developers can integrate custom functions with the sandbox
- Tokens can be refreshed without losing sandbox state
- Code examples available in JavaScript, Python, and cURL
Interesting facts
- The Gemini Interactions API documentation includes practical code examples.
- These enhancements aim to significantly streamline developer workflows.
What changed