Anthropic launches managed infrastructure for autonomous AI agents

Until now, teams shipping production-ready agents had to build their own secure containers, permission systems, and agent loops from scratch. Managed Agents handles all of that. According to Anthropic's documentation, the system provides an orchestration harness that independently calls tools, manages context, and handles errors. Anthropic claims this cuts the time from prototype to production by a factor of ten.

Sessions can run autonomously for hours, and results persist even if the connection drops. Built-in tools include bash commands, file operations, web search, and connections to external services via MCP servers. A separate "Research Preview" that isn't publicly available yet lets agents spin up other agents, coordinate parallel tasks, evaluate outputs, and manage memory. Those interested can sign up for a waitlist to access these features.

Early enterprise customers are on board, but only on Anthropic's infrastructure

Several companies are already using the system, according to Anthropic. Notion lets teams delegate tasks to Claude directly within their workspace. Rakuten built enterprise agents for sales, marketing, and finance that plug into Slack and Teams - each reportedly up and running within a week. Sentry paired its debugging agent with a Claude agent that writes patches and opens pull requests.

Managed Agents is available to all API accounts and requires a specific beta header that the SDK sets automatically. Anthropic notes that behavior may change between releases. Pricing is usage-based: standard token rates apply, plus $0.08 per session hour.

For now, the system runs exclusively on Anthropic's own infrastructure. The announcement doesn't mention whether it will become available through partner platforms like Amazon Bedrock or Google Vertex AI. For companies with multi-cloud strategies, that could be a significant limitation.