Developers shipping production agents on Claude can skip building hosting themselves, paying token rates plus $0.08 per session-hour.
Kenning's take
What it does
Claude Managed Agents is a set of composable APIs on the Claude Platform for building and deploying cloud-hosted agents. Later updates added memory, 'dreaming' that reviews past sessions to improve agents, outcomes that let Claude iterate towards a quality bar, and multi-agent orchestration. Usage costs standard token rates plus a runtime fee per session-hour.
- Hosted agent runtime and sandbox
- Memory across sessions
- Dreaming (research preview) to learn from past sessions
- Outcomes with success criteria
- Multi-agent orchestration and webhooks
Who it suits. Developers building production AI agents on Claude.
Key facts
- Category
- Developer tools
- Part of
- Claude
- Made by
- Anthropic
- Based in
- United States
- Released
- 8 April 2026
- How to get it
- public beta on the Claude Platform (API); dreaming by request in research preview
- Works on
- API
- In the UK
- Yes
- API price
- model token rates plus $0.08 per session-hour of active runtime
FAQ
How much do Claude Managed Agents cost?
Standard token rates for the model used, plus $0.08 per session-hour of active runtime.
When did Claude Managed Agents launch?
In public beta on 8 April 2026. Dreaming, outcomes and multi-agent orchestration followed on 19 May 2026.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
Claude Code
Anthropic's coding agent that reads a codebase, edits files and runs commands from the terminal, IDE or web
Claude Code Review
A Claude Code service that sends several agents to review each GitHub pull request and comment on problems
Claude Security
Anthropic's code scanner that finds vulnerabilities in a GitHub repository and suggests patches

Claude Keychain
A free Mac menu bar app that keeps API keys in the macOS Keychain so Claude Code can use them without seeing them.