Anthropic Just Launched Claude Managed Agents — Hosted Agent Infra at $0.08 an Hour

TamilTech AI summary

Anthropic just launched Claude Managed Agents, a fully hosted agent runtime that lets you describe a task via one API call while they handle the hard production plumbing like sandboxes, credential vaults, checkpointing, tracing, scoped permissions, and built-in tools. You pay about $0.08 per session-hour plus normal Claude token costs, which is a huge cut versus self-hosting equivalent infra that can run tens of thousands of rupees a month before anyone even runs an agent. It is still in beta, so every request needs the header `anthropic-beta: managed-agents-2026-04-01`, and features like Outcomes, multi-agent coordination, and persistent Memory remain in research preview. Early enterprise users such as Notion, Rakuten, and Asana are already on it for real workloads, and it feels more ergonomic for long autonomous tasks than Bedrock Agents, Vertex AI Agent Builder, or the Assistants API. If you are starting a new agent project, this is a strong default for most teams—just keep a fallback handy given recent outages, and skip it only if you have strict data-residency needs or a deep self-hosted setup already.

• What is Claude Managed Agents and when did it launch?
• How much does Claude Managed Agents cost?
• Which beta header do I need to access Claude Managed Agents?
• Which enterprise companies are already using Claude Managed Agents?

AI-assisted summary, checked by the TamilTech editorial team.