At Google I/O, the company introduced Gemini API Managed Agents — a service that promises to condense weeks of agent deployment work into a single API call. It’s also a sign that Google believes its ecosystem is ready to fully own the implementation layer, including the newly launched Antigravity CLI.
Before a single agent is written, teams already spend days on the eye-candy: getting execution environments up and running, managing sandboxes, wiring tool call infrastructure. Model providers like Anthropic have launched platforms to handle much of this work – but Google’s approach is different.
Google said blog post that managed agents in the Gemini API “remove complexity so you can focus on your product experience and agent behavior.” The service is available in preview via new custom templates in Google AI Studio.
The growth raised a real architectural question: should agent management live at the execution layer embedded in the model or its wrapper, or at the infrastructure layer as a separate runtime?
A comparison of the Google approach
Until recently, agent orchestration relied on frameworks that sat on top of the model, routing agents and allowing teams to separately control routing and execution. That layer is now absorbed by the platforms themselves.
Latest platforms As an orchestra into Claude Managed Agents in the model layer, not in a separate runtime platform. The idea is that the model owns the reasoning and orchestration layers, and the entities control the execution.
Through new capabilities in AWS, Bedrock AgentCore, adds controlled trailers combines initial tasks to deploy agents. Google’s approach goes even further, optimizing the model, trailers, and sandbox, and running everything in secure Google-managed environments.
Ramp’s Rene Sultan, who was cited in Google’s announcement, said the change was concrete: "The real change with Gemini Managed Agents is that agent runtime moves to the platform. With the sandbox, infrastructure, and execution cycle managed for you, developers can focus on producing the agent’s domain-specific behavior and iterating at a completely different pace."
A new orchestral reality
Businesses just starting out with agents may find the platform offerings from Anthropic and Google powerful, especially since they take the hassle out of deploying agents while still retaining some control. However, Google is pushing for a more vertically integrated system, while Anthropic is betting on the model layer as an orchestration plane and AWS is focusing on authorization.
And that comes with some risks, according to Arie Trouw, founder and CEO of XYO.
“An additional risk is that developers will deprecate previously defined services, which may now have unintended consequences for users at best or data breaches at worst,” Trouw told VentureBeat in an email. “It’s a classic example of the amazing hammer and everything starts to look like nails. I’ve seen this pattern many times over the past few decades as a developer and business builder.”
