What is Anthropic Managed Agents
Forgent relies on an Anthropic building block called Managed Agents. If you want to understand concretely what happens when you click “Deploy”, this page is for you. A 2-minute read.
The idea in one sentence
Managed Agents is the Anthropic API that runs your agent for you, on their infra. You describe what it should do, you send it over to them, and Anthropic takes care of running it, scheduling it, and monitoring it.
Without Managed Agents — the classic path
To run an AI agent in production, you usually need:
- A server (VM, container, or serverless function)
- Code to orchestrate the calls to Claude
- A scheduler (cron, queue, webhook trigger)
- A logging, monitoring, and alerting system
- A security layer for the tools (who can do what)
- Error handling, retries, and recovery from incidents
In short, several days of setup before the first line of business logic. And after that, you have to maintain all of it.
With Managed Agents — the short path
You send Anthropic what's called a blueprint — the spec sheet for your agent: its name, the Claude model it uses, the tools it has access to, the instructions it should follow. Anthropic registers this blueprint and hands you back a unique identifier. From there, they handle everything:
- Hosting — the agent lives on Anthropic's infra, and you have no server to manage.
- Execution — you launch the agent from your dashboard or via their API, and they run it.
- Sandboxed tools — each tool (bash, files, web fetch) runs in a secure, isolated environment.
- Logs and monitoring — accessible in the Anthropic console.
And where does Forgent fit in?
The catch with Managed Agents is that to use it well you need to know how to write a correct blueprint: pick the right Claude model, write precise instructions, declare the right tools in the right technical format. Not insurmountable, but not trivial either.
Forgent is the interface that writes that blueprint for you, from a simple plain-language description. You describe your need, the Builder guides you through 5 steps, we build the complete blueprint, and we send it to Managed Agents in one click.
Beta status
At the time of writing (May 2026), Managed Agents is in public beta. Anthropic regularly adds new models, tools, and MCP integrations. Forgent tracks these changes — check the changelog for the updates.
Going further
If you want the technical details (an agent's lifecycle, session management, what Anthropic handles vs. what stays on your side), head over to Concepts > Managed Agents. Otherwise, next step: Get started to get your hands dirty.