Anthropic Launches Claude Managed Agents to Build and Deploy AI Agents at Scale
Anthropic just made it dramatically easier to build AI agents. The company launched Claude Managed Agents in public beta — a fully managed platform that handles the heavy lifting of running autonomous AI agents in the cloud, so developers can focus on what their agents actually do rather than how they run.
What Are Claude Managed Agents?
Think of Managed Agents as Anthropic hosting and orchestrating your AI agents for you. Instead of building your own agent loop, tool execution framework, error recovery, and runtime infrastructure, you define the agent's task, tools, and guardrails — and Anthropic's platform handles everything else.
Each agent runs in a secure sandboxed container with pre-installed packages (Python, Node.js, Go), configurable network access rules, and mounted files. The agent can read files, run commands, browse the web, and execute code — all within Anthropic's managed infrastructure.
Key Features
- Secure sandboxed execution: Each agent runs in its own isolated container with scoped permissions and network access rules
- Tool orchestration: Built-in tools for bash commands, file operations, web search, and more via the
agent_toolset_20260401tool type - Checkpointing: Agents can pause and resume, maintaining state across long-running tasks
- Multi-agent pipelines: Support for complex workflows where multiple agents collaborate on different parts of a task
- Authentication and permissions: Scoped access controls for what each agent can do
- Context management: Built-in prompt caching, compaction, and performance optimizations
- Error recovery: Automatic handling of failures and retries
From Months to Weeks
Anthropic says the biggest value proposition is speed. Building a production-quality agent system from scratch typically takes months — you need to handle tool execution, state management, error recovery, security sandboxing, scaling, and monitoring. Managed Agents compresses this to weeks or even days.
The platform supports both simple single-task flows and complex multi-agent systems. A developer can define an agent with a model, system prompt, tools, MCP servers, and skills, create it once, and reference it by ID across sessions.
How Much Does It Cost?
Pricing follows a consumption model: standard Claude token rates apply, plus $0.08 per session-hour for active runtime. This means you only pay for the compute time your agents actually use — no idle infrastructure costs.
The trend toward enterprise AI consolidation makes this pricing model attractive for companies that want to experiment with agents without committing to massive infrastructure investments.
Who's Already Using It?
The early adopter list includes some heavy hitters: Notion, Rakuten, and Asana are among the first companies building on Managed Agents. These are productivity and enterprise companies that have natural use cases for autonomous AI workflows — from automated project management to intelligent document processing.
The Agent Platform Wars
This launch positions Anthropic squarely against OpenAI's Codex and Google's agent offerings in the emerging "agent platform" category. The AI talent war extends to agent infrastructure, where the company that makes it easiest to build and deploy agents could capture the next wave of enterprise AI spending.
What makes Managed Agents distinctive is its focus on security and control. While other platforms emphasize capability, Anthropic is betting that enterprises will prioritize sandboxed execution, scoped permissions, and checkpointing — features that matter when AI agents are handling sensitive data and critical workflows.
Getting Started
Managed Agents is available now in public beta through the Claude Platform. Developers can use the Claude API, the new CLI, or build with Claude Code to deploy their first agent. Some features including outcomes, multiagent coordination, and persistent memory are currently in research preview.
Frequently Asked Questions
What are Claude Managed Agents?
Claude Managed Agents is Anthropic's fully managed platform for building and deploying autonomous AI agents. It provides secure sandboxed execution, tool orchestration, checkpointing, and multi-agent pipelines so developers can focus on agent design rather than infrastructure.
How much does Claude Managed Agents cost?
Managed Agents uses consumption-based pricing: standard Claude token rates plus $0.08 per session-hour for active runtime. You only pay for compute time your agents actually use.
Which companies are using Claude Managed Agents?
Early adopters include Notion, Rakuten, and Asana. These enterprise companies are using Managed Agents for automated workflows, intelligent document processing, and AI-powered project management.
How is this different from building your own AI agent?
Building a production agent system from scratch requires months of work on tool execution, state management, error recovery, security, and scaling. Managed Agents handles all of this infrastructure, letting developers go from concept to production in days or weeks instead of months.