Where to Host Completed Agents? Google AI Studio is Preparing a Managed Hub

Google’s AI development tool AI Studio is reported to be developing a new hub called Managed Agents. According to Alexey Shabanov, author of TestingCatalog, the feature is already in preparation, but Google hasn’t officially announced it yet, and the launch time and specific capabilities may still change.

AI Studio is Google’s development platform for the Gemini model. Developers can write prompts, upload files, and connect tools to quickly build an agent that can chat or perform multi-step tasks. Such tools are convenient for demos, but when moving to actual use, small issues arise: where the agent is deployed, who can access it, which tools are called, how to roll back versions, and where to check logs and costs.

The managed agent hub is designed to fill this gap. Inferred from its name, it may provide a centralized interface for developers to manage agents running on Google’s infrastructure, reducing the cost of maintaining servers and scheduling tasks themselves. For people with development backgrounds, this relates to agent lifecycle management; for regular users, this entry may not be visible in the short term, but it will affect the stability and distribution of agents in the Gemini ecosystem in the future.

TestingCatalog often reveals new features through product interfaces and code clues before official releases, so this kind of information can be seen as a confirmation of direction, not a complete release note. There is currently no evidence that it will be a consumer-facing agent store; it is more accurate to understand it as an agent console for developers.

This also aligns with the common path taken by major tech companies in developing agents. Model capabilities are gradually becoming a baseline, and tool calling, permission control, state maintenance, and deployment environments are starting to become more practical differentiators. Google already has development tools like Gemini API and Agent Development Kit, and adding the managed hub to AI Studio is equivalent to further shortening the path from prototype to online operation.

What is worth watching next is how many operation and maintenance details Google will include in this hub, especially logs, versions, permissions, and trace chain tracking. These functions are not flashy, but they determine whether an agent can be stably used by others.

Source: TestingCatalog