Google Deepmind has introduced four new capabilities to its Gemini API managed agents, enhancing their functionality for developers. These updates allow agents to run asynchronously in the background without requiring an open HTTP connection. Additionally, remote MCP servers can now be linked directly to internal databases or APIs. Developers can also integrate custom functions with built-in sandbox tools, and credentials like tokens can be refreshed between interactions without disrupting the sandbox state. All these features are accessible through the Gemini Interactions API, with code examples provided for JavaScript, Python, and cURL in the documentation. Source: thedecoder
The new background execution feature enables agents to operate independently, reducing the need for continuous connections. This allows for more efficient resource management and improved performance in long-running tasks. The addition of MCP support facilitates direct communication between remote servers and internal systems, streamlining data flow and integration. Developers can now leverage custom functions alongside sandbox tools, offering greater flexibility in application design. Credential management has also been enhanced, allowing tokens to be refreshed without losing the sandbox state, ensuring continuity in agent operations. Source: thedecoder
The updates are part of Google Deepmind's ongoing efforts to improve the Gemini API, providing developers with more tools to build advanced applications. Source: thedecoder*
