Claude Managed Agents now plans runs of up to 1,000 agents
Anthropic on October 9 put dynamic workflows, a new kind of multiagent orchestration in Claude Managed Agents, into public beta, its ClaudeDevs account announced on X. An agent configured with the multiagent type multiagent_20261001 starts one when asked to run a workflow, and Claude then writes the plan and orchestrates up to 1,000 agents per run. These workflows can use a lot of tokens, which is why Anthropic suggests starting with a scoped task and raising the complexity step by step.
That plan runs across the agents in phases, and the results are combined at the end. Multiagent orchestration as such joined the Managed Agents public beta on May 6, with a lead agent handing pieces of a job to specialists that each have their own model, prompt and tools, as the Claude blog put it then.
In Claude Code, where dynamic workflows first appeared as a research preview on May 28, Anthropic’s post described scripts that run “tens to hundreds of parallel subagents in a single session”; the Managed Agents cap is 1,000 per run.
Onboarding in Claude Code starts with /claude-api managed-agents-onboard bug-hunter, and templates are in the agent quickstart on the Claude Platform.
Up next
OpenAI says fired safety trio broke sensitive-info rules
“A significant breach of trust”: that’s OpenAI’s public reason for firing safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni. The company insists their safety warnings had nothing to do with it.
Related stories
- Claude API sends its traces to your observability tool
- Anthropic’s daily-brief agent treats a failed read as news
- Claude Managed Agents get auto mode and a session viewer
- Claude apps switch to auto mode by default next week
- Claude Duet works while you’re away and asks before acting
- Claude Marketplace opens its shelf to outside builders
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

