Building the AI part of an agent takes days, and infrastructure takes months.
Most of that work, like sandboxing, state, permissions, and error handling, is invisible but time-consuming, and many teams get stuck here.
Anthropic solves this with Claude Managed Agents. It offers APIs to deploy cloud-hosted agents at scale, now in public beta.
Teams can skip heavy infrastructure work and start deploying faster.
What is Claude Managed Agents?
Claude Managed Agents is a set of composable APIs that handle the infrastructure side of running AI agents. You define what the agent should do, what tools it has access to, and what success looks like. Claude handles the rest.
Under the hood, it includes secure sandboxed execution, built-in authentication, context management, and session tracing. You don't have to wire any of that up yourself.
It also supports long-running sessions. Likewise, agents can work autonomously for hours, with outputs persisting even through disconnections. For teams building anything beyond simple prompt-response flows, that's a real difference.
Why teams are reconsidering AI agent deployment
The promise isn't theoretical. In internal testing, Managed Agents improved task success rates by up to 10 points over standard prompting loops, with the biggest gains on harder, longer tasks.
Teams already using it have shipped in days rather than months. A few examples from the announcement:
- Notion uses it to let teams delegate open-ended work to Claude directly inside their workspace from writing code to generating presentations. Multiple tasks can run in parallel while the team collaborates on results.
- Rakuten deployed specialist agents across engineering, product, sales, and finance, each within a week. Employees assign tasks in Slack or Teams and get back actual deliverables like spreadsheets and slides.
- Sentry paired with their existing debugging agent with a Claude-powered agent that writes the fix and opens the pull request. The whole integration shipped in weeks, not months.
- Asana used it to add advanced features to their AI teammates product shipping capabilities dramatically faster than they could have with custom infrastructure.
The pattern across all of them is the same: less time on infrastructure, more time on the product.
Key features of Claude Managed Agents
- Production-grade infrastructure out of the box- Secure sandboxing, authentication, and tool execution is handled for you so there is no need to build or maintain custom runtime systems
- Long-running, persistent sessions- Agents can run autonomously for hours while retaining state which makes them suitable for tasks that require multiple steps over time
- Multi-agent coordination- Agents can create and manage other agents to handle complex tasks in parallel, this feature is in research preview with early access available
- Scoped permissions and execution tracing- Agents can connect to real systems with built in identity control and fine-grained permissions while every action and tool call is logged in the Claude Console
- Self-evaluation mode- You define the goals and success criteria, and Claude continues to iterate until those outcomes are achieved, this is available in research preview along with multi agent coordination
Before starting to build agents, getting the architecture right is key. GoML’s AI System Design whitepaper details essential design choices for scalable, production-ready AI systems essential reading before deploying your first agent.
How pricing works
Managed Agents is priced on consumption. Standard Claude Platform token rates apply, plus $0.08 per session-hour for active runtime.
Getting started
Claude Managed Agents is now available. You can begin with the Claude Console, explore the documentation at platform.claude.com, or use the CLI to deploy your first agent. Claude Code users can also get started by prompting, "start onboarding for managed agents in Claude API."
If you're exploring faster ways to build and launch agents, check out GoML’s AI Matic page. It provides LLM accelerators and frameworks to accelerate agent development, deployment, and scaling for enterprise teams skipping months of infrastructure work.
The infrastructure challenge has slowed teams for a long time. Managed Agents removes that barrier. If deployment has been blocking your progress, this is a practical place to start.
