Google has expanded Managed Agents in Gemini API with several new features designed to help developers build more reliable AI agents. Overall, the update focuses on practical improvements for real-world use. In particular, it adds support for background execution for async interactions, easy connection to remote MCP servers, custom functions, and credential refresh.
Takeaways 🚀
✅Managed Agents gain async background support
✅Remote MCP connections become easier
✅Custom functions extend agent workflows
✅Credential refresh supports longer sessions
✅Gemini API targets production reliability
At its core, Managed Agents in Gemini API simplify agent development. Instead of requiring developers to manage every technical step manually, the platform handles reasoning, code execution, package installation, file management, and web access inside a secure cloud sandbox. As a result, teams can spend less time on infrastructure and more time building useful applications.
Background execution for async interactions
First, one of the most important additions is background execution for async interactions. In many cases, long-running tasks can become difficult to manage when an application must keep an HTTP connection open for the full process. Because of that, these workflows can become fragile or inefficient.
With background execution, however, developers can start a task and receive an interaction ID immediately. The task then continues on the server without blocking the application. After that, developers can check the status later, stream progress updates, or reconnect when the work is complete.
As a result, this feature makes Managed Agents in Gemini API better suited for tasks that take more time. For example, it can support code analysis, large research workflows, and other processing jobs that may not finish quickly.
Easy connection to remote MCP servers
In addition, Google is introducing easy connection to remote MCP servers. MCP, which stands for Model Context Protocol, helps agents connect with external tools and systems.
Previously, developers often needed custom middleware or proxy layers to connect agents to internal APIs, databases, or private systems. Now, by contrast, managed agents can connect directly to remote MCP servers. Because of this, integration becomes simpler and more direct.
Just as importantly, developers can combine remote MCP tools with built-in capabilities such as Google Search and code execution. Therefore, they can create more flexible and connected agent workflows.
Support for custom functions
Beyond that, the update adds support for custom functions. This feature gives developers a way to combine Gemini’s built-in tools with their own business logic.
For instance, the agent can complete built-in server-side actions automatically. Then, when it reaches a step that requires a custom function, the interaction can pause so the client can run that function and return the result.
In this way, developers gain more control over specialized tasks. At the same time, they can keep using the platform’s managed capabilities for the rest of the workflow. Consequently, this balance can make it easier to connect agents to internal systems and company-specific processes.
Credential refresh for longer workflows
Meanwhile, Google is also adding credential refresh, which addresses a common challenge in longer or multi-step interactions. Over time, access tokens and API keys can expire. When that happens, workflows may fail or require a restart.
With credential refresh, developers can update network credentials in a later interaction while keeping the same environment active. In other words, the system preserves the existing filesystem state, installed packages, and cloned repositories, even while it replaces the previous network rules with updated credentials.
As a result, long-running workflows become easier to maintain. Instead of rebuilding everything from the beginning, teams can continue working within the same environment.
Why this update matters
Taken together, these improvements make Managed Agents in Gemini API more practical for production use. Rather than focusing only on model intelligence, Google is also improving the operational side of agent development.
More specifically, background execution for async interactions supports long-running tasks, easy connection to remote MCP servers simplifies external integrations, custom functions allow developers to add local logic, and credential refresh helps maintain secure access over time.
Because of these changes, developers may spend less time handling technical setup and more time creating meaningful applications. Likewise, organizations may find it easier to connect AI agents to the systems and workflows they already use.
Final thoughts
In conclusion, the expansion of Managed Agents in Gemini API marks a meaningful step forward for practical AI development. The update does not just add new features. Rather, it improves how agent systems operate in real development environments.
By supporting background execution for async interactions, easy connection to remote MCP servers, custom functions, and credential refresh, Google is making managed agents more flexible, reliable, and easier to integrate. Ultimately, that makes the platform more useful for teams that want AI agents that can do more than simply respond — they can also work effectively within real applications and systems.
