Anthropic Adds Dynamic Workflows to Claude Managed Agents, Supporting 1,000-Way Parallelism

On October 9, Anthropic announced dynamic workflows for Claude Managed Agents. In this mode, a main agent plans tasks, assigns subtasks to subagents, and aggregates results after completion. Each execution can coordinate up to 1,000 AI agents in parallel.

To demonstrate the capability, Anthropic tested code vulnerability detection by hiding 70 bugs in a codebase of 116,000 lines. The dynamic workflow consistently detected 66 bugs, while a single agent found only 14 to 27 per run. The approach is suitable for breaking large tasks like code review into parallel subtasks.