HomeAIGoogle Expands Managed Agents in Gemin
AI

Google Expands Managed Agents in Gemini API

Google DeepMind added background execution, remote MCP server support, custom functions, and credential refreshing to the Gemini Interactions API.

WHAT YOU NEED TO KNOW
  • Google introduced background task execution in the Gemini API using an asynchronous interaction ID system.
  • Managed agents can now connect directly to remote Model Context Protocol (MCP) servers alongside sandbox tools.
  • Developers can refresh expiring network credentials on existing sandboxes without losing filesystem state.
  • Custom function calling uses step matching to trigger client-side execution when needed.

Google announced new capabilities for Managed Agents in the Gemini API on Jul 07, 2026, introducing background execution, remote Model Context Protocol server integration, custom function calling, and credential refreshing across interactions. The company said the updates address developer feedback to help build production-ready autonomous agents.

Through the Gemini Interactions API, managed agents operate inside an isolated cloud sandbox to process reasoning, execute code, install packages, manage files, and fetch web information via a single endpoint call. To resolve issues with HTTP connections during long-running operations, Google added background execution support. Developers can pass a background parameter set to true, causing the API to return an interaction ID immediately so client applications can poll status, stream progress, or reconnect later while the task finishes remotely on the server.

Tools and custom execution

Google added direct connections to remote Model Context Protocol (MCP) servers, allowing managed agents to access internal APIs or private databases without custom proxy middleware. Developers can combine an MCP server tool alongside built-in sandbox capabilities, such as Google Search and code execution, within the same interaction.

The Gemini API now supports custom function calling alongside sandbox tools using step matching. Built-in tools execute automatically on Google servers, whereas custom domain functions transition the interaction to require action from the client, allowing local business logic to run before sending results back to the agent in subsequent turns.

Credential refresh and setup

Google introduced network credential refreshing to handle short-lived API keys and expiring access tokens. Developers can rotate keys or refresh credentials by sending an existing environment ID paired with a new network configuration on a subsequent interaction. The sandbox retains its filesystem state, cloned repositories, and installed packages while updating the access rules immediately.

Code examples for the new features rely on the `@google/genai` JavaScript SDK. For AI coding agents, Google made the skill available via npx skills add google-gemini/gemini-skills --skill gemini-interactions-api, while Python and cURL implementations are located in the Antigravity agent documentation.

Xentir Media
Xentir Media NewsroomSource-backed AI and technology coverage, drafted by Xentir's automated editorial system under fixed human-set rules. See our editorial policy and AI usage policy.
J
Jomon · Founder & EditorFounder and editor of Xentir Media. Sets the editorial rules the newsroom system runs under, and is accountable for its corrections. About Jomon · [email protected]
The Xentir Brief
The developments worth knowing — one useful email.
Get the Brief →