title: "Google expands Gemini API managed agents with new infrastructure" slug: "google-expands-gemini-api-managed-agents-with-new-infrastructure" published: "2026-07-22" beat: "Launches" tags: ["Launches", "Tools"] creator: "Agentry Newsroom" editor: "Susanne Sperling, Editor — Human in the Loop" tools: ["Claude (Anthropic)", "Perplexity Sonar"] creativeWorkStatus: "verified" dateReviewed: "2026-07-22" aiActArticle50: "compliant" humanView: "https://agentry.news/google-expands-gemini-api-managed-agents-with-new-infrastructure" agentView: "https://agentry.news/agent/google-expands-gemini-api-managed-agents-with-new-infrastructure"
Google announced new capabilities for Managed Agents in its Gemini API on July 22, 2026, adding background execution, remote server integration, custom function calling, and credential refresh to help
Drafted by an AI agent. Verified by Susanne Sperling, Editor — Human in the Loop. AI policy.
Google expanded Managed Agents in the Gemini API with four new infrastructure capabilities designed to increase the sophistication and autonomy of agent deployments Google.
The update adds background execution for asynchronous task handling, remote MCP (Model Context Protocol) server integration, custom function calling, and credential refresh and rotation across interactions. Google's Managed Agents let developers invoke agents through a single API endpoint while the platform handles reasoning, code execution, package installation, file management, and web retrieval inside an isolated cloud sandbox Google.
Background execution enables agents to continue working on long-running or deferred tasks without blocking the client. Remote MCP server support lets developers connect external tools and data sources to agent sandboxes, expanding what agents can access and act upon. Custom function calling allows developers to define domain-specific operations that agents can invoke as part of their reasoning and action loops.
Credential management now supports refresh and rotation: developers can pass an existing environment_id with a new network configuration on subsequent interactions, enabling agents to update API keys, tokens, and authentication without requiring sandbox restart Google.
These features address operational constraints that have limited agent adoption in production environments. Background execution removes the latency penalty for asynchronous workflows. Remote MCP integration standardizes how agents connect to external systems, reducing custom integration work. Credential refresh eliminates a common operational friction point — rotating secrets without agent downtime.
The timing reflects competitive pressure in the managed agent space. OpenAI, Anthropic, and other platforms have released or expanded agent offerings over the past 12 months. Google's emphasis on isolated sandboxing, package installation, and web access differentiates its model, and these infrastructure additions extend that positioning by enabling stateful, long-running, and externally integrated agent operations.
The Managed Agents feature remains in general availability, with these capabilities rolling out to developers in the Gemini API console and SDKs.