Anthropic’s managed agents in the wild
Anthropic’s Claude Managed Agents are being used in production workflows to handle state management and error recovery, with deployments reported at Notion and Asana for rapid prototyping to production. The platform is being highlighted as a practical way to build narrow, ownership‑bounded agents that orchestrate tasks rather than act as general chatbots (x.com).
Anthropic’s Claude Managed Agents are running in production at Notion and Asana, handling stateful workflows and automated error recovery. (claude.com) Anthropic announced Claude Managed Agents on April 8, 2026 and said the service is available in public beta on the Claude platform. (claude.com) The company markets the product as an end‑to‑end, cloud‑hosted suite that shortens the path from prototype to production, claiming “get to production 10x faster” in its launch post. (claude.com) Notion, Asana, Rakuten and Sentry are listed by Anthropic and multiple news outlets as early adopters already using Managed Agents for real customer workflows. (claude.com) Anthropic’s documentation says Managed Agents provide long‑running sessions, memory, tool integrations, prompt caching and a secure sandbox to run commands and access files. (platform.claude.com) Industry coverage and Anthropic’s engineering blog frame the product as a hosted “harness” that takes on infrastructure tasks—logging, orchestration and runtime—so engineering teams don’t have to rebuild those layers. (anthropic.com) Security researchers warn that hosted agent platforms expand the attack surface: Ship Safe published a 12‑rule scanner for Managed Agents misconfigurations, and analysts flag vendor lock‑in as a strategic risk for enterprises. (shipsafecli.com) Anthropic says Managed Agents are in public beta and available to customers now; enterprises such as Notion and Asana will serve as early production tests while the company adds enterprise controls like role‑based access and observability. (claude.com)
