Claude Code Now Writes Its Own Agent Orchestration
Dynamic Workflows just shipped in research preview — Claude generates its own orchestration scripts on the fly, runs subtasks in parallel, and verifies results before surfacing them. 86% of teams were already running agents in production when this landed.
NeuroX AI · June 2, 2026
Anthropic shipped Dynamic Workflows to Claude Code this week, and 86% of engineering teams were already deploying agents in production when it landed. The feature closes the last manual bottleneck in agentic work: writing the orchestration logic itself.
The old workflow: engineer defines which agents run, in what order, what they check. The new workflow: hand Claude an objective and it generates the coordination script on the fly — breaks work into subtasks, runs them in parallel, compares results, and iterates until they converge. Interrupted runs resume from where they stopped, which matters for jobs that span hours.
The use cases Anthropic highlighted are the ones teams already pay engineers to manage manually: security audits across a codebase, large-scale migrations, widespread bug investigations. The same tasks where "I ran Claude on it" always ended with "I still had to stitch everything together."
Two constraints to budget for:
- Token consumption runs substantially higher than a standard session — cost monitoring is not optional
- Start with well-scoped tasks before scaling to the full repo
Available now in research preview for Max, Team, and Enterprise plans, and through the API on Bedrock, Vertex AI, and Foundry.