Anthropic launched Claude Managed Agents, a new service for developers building and deploying AI agents - software that completes multi-step tasks autonomously rather than responding to single prompts.
The service sits between the raw Claude API and fully custom agent infrastructure. Instead of writing all the orchestration code yourself (the logic that decides what an AI agent does next, how it handles failures, and how it maintains state across a long task), Managed Agents handles that layer.
This is the right product for this moment. Agent-based workflows moved from demos to actual production use over the past year. The friction point shifted from "can this work?" to "how do I keep this working reliably?" - covering retry logic, session persistence, monitoring, and cost control. Managed infrastructure solves that without requiring developers to build it from scratch.
Anthropicking is competing directly with OpenAI's Assistants API and third-party orchestration tools like LangChain and CrewAI. The differentiator would need to be Claude's reasoning quality combined with lower operational overhead - a case Anthropic is well-positioned to make given Claude's consistent performance on complex, multi-step instructions.
Pricing and full capability details are still being confirmed. For developers already building on Claude via API, this is the logical next step before committing to custom agent infrastructure.