By Ken Muthomi - Reporter
U.S July 8, 2026 - Google has launched a series of new capabilities and features for its Managed Agents in the Gemini API, which are designed to help developers build more reliable AI agents.
According to the tech giant, the upgrades include background execution, remote Model Context Protocol (MCP) server integration, custom function calling, and credential refresh across interactions.
Google has said that the managed agents in the Gemini Interactions API will allow developers to call a single endpoint while Gekini handles reasoning, code execution, package installation, file management, and web information retrieval inside a cloud sandbox.
“These updates directly address developer feedback and product needs so you can build reliable, production-ready agents,” Google said.
A key addition is te background execution, through which developers will have the ability to run long-running tasks independently without maintaining an active HTTP connection.
By enabling the background: true option, the API immediately returns an interaction ID that applications can use to monitor progress, reconnect later, or retrieve results once the task has been completed remotely.
The company has also introduced remote MCP server integration, which will allow managed agents to connect directly to remote Model Context Protocol servers, erasing the need for developers to build custom proxy middleware for private databases or internal APIs.
“You can mix and match remote tools with built-in sandbox capabilities. Pass an mcp_server tool at interaction time alongside Google Search or code execution to let the agent communicate with your endpoints from its secure sandbox,” the company stated.
Another enhancement is custom function calling, which lets developers integrate their own business logic alongside Gemini's built-in sandbox tools.
Google has also introduced network credential refresh, which is set to address a common challenge in AI deployment. This will allow developers to update expired access tokens or rotate API keys without recreating the AI agent's working environment.
“These updates turn managed agents into asynchronous workers that operate inside real development environments without blocking your application,” the company stated.
More from Kenya