Claude can operate a maximum of 1,000 agents at the same time
A test by Anthropic found 66 of 70 bugs with the new workflow, but no one knows if this holds for other tasks.
Claimed, not confirmed
Anthropic added dynamic workflows to Claude Managed Agents. A lead agent divides the work into tasks, gives the tasks to sub-agents, and then makes one result from their results. A maximum of 1,000 agents can operate at the same time in each execution. In a test by Anthropic, one agent found 14 to 27 of 70 bugs, but the workflow found 66. A senior OpenAI engineer thinks that agent swarms use too many tokens.
How it works
The lead agent divides the work into tasks and gives the tasks to sub-agents. When the sub-agents finish, the lead agent makes one result from their results. Claude Managed Agents is the Anthropic platform for agents. Dynamic workflows are new on this platform.
The Anthropic test
Anthropic put 70 bugs in a large program. One agent found 14 to 27 of the bugs in each test. The workflow found 66 of the bugs in each test.
The cost question
A senior OpenAI engineer thinks that agent swarms use too many tokens. Anthropic agrees that a workflow can use many tokens. It tells users to start with a small task. It is not known if the result holds for other types of task.
How to start
- Select the agent type
multiagent_20261001. - Read the documentation.
- Use the command
/claude-api managed-agents-onboardin Claude Code.
This is a brief. We point to the report and do not rewrite it. Read it at the source below.
Sources
Posted
