Google's Gemini Managed Agents Get a Quieter Upgrade: Faster Default Model, Guardrails, and a Free Tier
Gemini 3.6 Flash is now the default for managed agents, and new environment hooks let developers block or audit what the AI does inside its sandbox.
Key points
- Google's Gemini API managed agents now default to the Gemini 3.6 Flash model, with no code changes required for existing users.
- New environment hooks let developers run custom scripts before or after any tool call the agent makes inside its cloud sandbox.
- A new
max_total_tokenssetting caps how much an agent can spend on a task before it safely pauses. - Managed agents are now available on free tier Google AI projects, so developers can try them without active billing.
- Scheduled triggers let agents run on a recurring timetable, reusing the same sandbox between runs.
Google has quietly upgraded the tools it offers developers who build AI agents, software that can carry out multi-step tasks like editing files or running code on its own. The changes, announced by Google AI on the DeepMind blog, focus less on raw model power and more on the boring but important work of keeping agents cheap, safe, and predictable.
Here is what actually changed, and why it matters for the apps ordinary people will end up using.
What is a managed agent, in plain words?
A managed agent is an AI worker that Google runs for you inside an isolated cloud sandbox, a small, walled-off computer in the cloud. You give it a task in one API call, and it reasons, writes code, installs packages, edits files, and browses the web to get the job done.
Developers use these to build things like automated code reviewers, research assistants, or the AI analyst tools used inside banks. The user just sees a helpful feature. The agent is doing the work behind the scenes.
What changed with the default model?
The antigravity-preview-05-2026 agent now runs on Gemini 3.6 Flash by default, replacing the older Gemini 3.5 Flash. Existing code picks it up automatically on the next call.
Google describes 3.6 Flash as a balanced model for reasoning, coding, and tool use. Developers who want to spend less can still pin the agent to Gemini 3.5 Flash-Lite, which Google says is the lowest cost and latency option in the 3.5 family.
| Model | Role |
|---|---|
| Gemini 3.6 Flash | New default, balanced |
| Gemini 3.5 Flash | Previous generation |
| Gemini 3.5 Flash-Lite | Lowest cost and latency |
Why do the new "hooks" matter?
Environment hooks let a developer run their own script before or after every action the agent takes inside its sandbox. That means a company can block risky tool calls, force code formatting, or audit what the AI did, without trusting the model to police itself.
A developer drops a .agents/hooks.json file into the environment. The runtime then fires their scripts on pre_tool_execution or post_tool_execution events. If a pre-execution script returns a deny decision, the tool call is skipped and the reason is passed back into the model's context so it knows to try something else.
Alston Lin, founder and CTO of AI-native investment bank OffDeal, said his team uses post-execution hooks to verify company logos inside investment banking decks: checking each one for the right company, aspect ratio, transparent background, and contrast, before it lands in a slide. Before hooks, Lin said, that validation had nowhere to run because the sandbox was remote.
For the rest of us, this is the difference between an AI feature that occasionally embarrasses a company and one that has a safety net.
What about runaway costs?
Agents that loop autonomously can burn through tokens, the small chunks of text an AI charges by, faster than a human notices. Google is addressing this with a new max_total_tokens cap developers can pass in when creating an interaction.
When the agent hits the limit, it pauses with a status of incomplete and preserves its environment. The developer can then resume it later with a fresh budget by passing the previous interaction ID. New scheduled triggers also let an agent run on a cron schedule, reusing its sandbox between runs so files persist.
Who can try it now?
Managed agents are now available on free tier projects, meaning developers can experiment using an API key from a Google AI project without active billing. There is also a new Environments API for listing and deleting sandbox sessions from code, rather than waiting for the default seven-day timeout.
None of this is a headline-grabbing model launch. It is the plumbing that decides whether the AI agents built on top of Gemini feel reliable or reckless.
