Anthropic just launched Claude Managed Agents, a cloud-hosted platform that handles the unglamorous infrastructure work of deploying AI agents at scale. The pitch is straightforward. Stop spending months building sandboxing, credential management, state persistence, and orchestration. Define your agent's tasks and guardrails, and Anthropic runs it on their infrastructure. It's available now in public beta.
Agent infrastructure has been a real pain point. Every team building production agents hits the same wall. Secure code execution. Long-running sessions. Permissioning. Tracing. That's months of engineering work before users see anything. Managed Agents abstracts all of it. In Anthropic's internal testing on structured file generation tasks, the platform improved task success by up to 10 points over standard prompting loops, with the biggest gains on the hardest problems.
Early adopters are already shipping. Notion integrated Managed Agents to let teams delegate complex tasks like coding and generating presentations without leaving their workspace, according to Product Manager Eric Liu. Rakuten deployed specialist agents across multiple business units in about a week each. Asana used it to build AI Teammates, their collaborative agents that work alongside humans in projects. Sentry paired their debugging agent with a Claude-powered patch writer. One flow, from flagged bug to reviewable fix.
The platform integrates with AWS Bedrock, Google Cloud Vertex AI, and Microsoft Foundry, which matters for enterprise adoption. Multi-agent coordination is available in research preview, letting agents spin up and direct other agents for parallel work. Anthropic is positioning itself as more than a model provider here. They're selling the whole stack: the models (Opus 4.6, Sonnet 4.6, Haiku 4.5), the orchestration, the infrastructure, the governance. Whether that lock-in is worth the speed gain is a question each team will have to answer for themselves.