Claude Managed Agents adds workflows for up to 1,000 agents
Illustration, not documentary evidence of the event.
Anthropic has added dynamic workflows to Claude Managed Agents, allowing a lead agent to plan work, assign tasks to sub-agents and combine their results. Each execution can run up to 1,000 agents in parallel, according to the-decoder.com, citing Anthropic.
The reported evidence is a bug-finding exercise from Anthropic’s own testing. In a 116,000-line codebase containing 70 deliberately hidden bugs, a single agent found between 14 and 27 per run. The dynamic workflow consistently found 66. That is a result from a specific company-run test, not independent verification of performance across other workloads.
For teams using agents to review code, the useful decision is whether splitting a review improves verified findings enough to justify the token use. Start with a bounded review task and run both a single agent and a small dynamic workflow against the same code and instructions. Check each reported issue, count unique confirmed findings and compare token consumption. Duplicate reports and unsupported findings should not count as additional coverage.
The orchestration mechanism matters as much as the parallel capacity. Because the lead agent assigns work and merges results, a useful trial should also examine that final synthesis: does it preserve valid findings from the sub-agents, resolve overlapping reports and give reviewers enough detail to act? This suggests evaluating the complete workflow rather than judging it only by how many agents ran.
The 1,000-agent ceiling is a capacity limit, not a recommended starting point. The report says Anthropic warns that dynamic workflows can consume substantial tokens and recommends starting small. Increase parallelism only when another trial produces enough additional useful results to support that choice. The supplied report does not establish the cost of those gains or whether they transfer to other task types.
To try the feature, the report specifies the agent type multiagent_20261001. It also identifies /claude-api managed-agents-onboard in Claude Code as an onboarding route. Define the review scope and acceptance criteria before expanding the workflow so the experiment answers a practical question about your own work.