Anthropic has launched Claude Managed Agents in public beta, a cloud-hosted API suite meant to speed up how companies build and run AI agents. The company said the platform can enable AI agent deployment up to 10 times faster by taking on the infrastructure work that often slows projects down.
The system is built to handle sandboxed execution environments, state management, credential handling and permission systems, and Anthropic said it provides production-grade sandboxing, authentication and tool execution out of the box. That matters because agents are meant to run on their own for long stretches, and sessions can persist through disconnections instead of stopping when a connection drops.
Anthropic is also testing features that push the model beyond single-agent tasks. Multi-agent coordination is in research preview and would let one agent spawn and direct others to parallelize work. A separate self-evaluation capability is also in research preview, allowing developers to define success criteria while Claude iterates toward meeting them.
Read Also: Google Stock Tied as Anthropic Signs Major TPU Deal With Google, Broadcom
The launch comes with early evidence that the system can improve performance on narrow tasks. In internal testing, Anthropic said success rates rose by as much as 10 percentage points on structured file generation tasks compared with standard prompting approaches. The company is pitching the platform as a way to strip out the months of setup that can delay agent-based AI projects before they ever reach users.
Several companies are already using variants of the system. Notion deployed Claude directly into workspaces through Custom Agents in private alpha, Rakuten stood up enterprise agents across product, sales, marketing, finance and HR within a week per deployment, and those agents plug into Slack and Teams. Asana built AI Teammates that work alongside humans inside project management workflows, while Sentry paired its Seer debugging agent with a Claude-powered counterpart that writes patches and opens pull requests.
Read Also: Project Glasswing: Anthropic opens Mythos Preview to security partners
Anthropic’s bet is that the next phase of enterprise AI will be decided less by model demos than by whether agents can be deployed, governed and monitored like real software. Session tracing and integration analytics in the Claude Console are part of that pitch, and the early adopters suggest the company is trying to prove Claude Managed Agents can move from experiment to infrastructure.
