Dynamic workflows in Claude Code

Claude breaks a problem into subtasks and fans them out across tens to hundreds of subagents running in parallel, dynamically writing orchestration scripts that run in a single session and checking its work before anything reaches you. It targets broad work like codebase-wide bug hunts and security audits and large migrations spanning thousands of files, verifying each result before folding it into the final answer. Available as a research preview on Max, Team, and Enterprise plans.

Claude Code Agents

Official announcement →

This article is a summary based on official documentation.

Overview

Anthropic introduced dynamic workflows in Claude Code: Claude breaks a problem into subtasks and fans them out across tens to hundreds of subagents running in parallel, dynamically writing the orchestration scripts that coordinate them in a single session and checking its work before anything reaches you.

Key features

Orchestrating parallel subagents

Large tasks ran as a single sequential pass, so broad work like a codebase-wide search was slow and easy to miss things in. Dynamic workflows let Claude break a problem into subtasks and fan them out across tens to hundreds of subagents running in parallel, dynamically writing the orchestration scripts that coordinate them in a single session.
Verification and convergence

Merging parallel results directly can let false positives or inconsistencies through. Workflows verify each result independently before folding it into the final answer, iterating until the answers align. For critical work, adversarial agents test the results before delivery.
Built for large-scale work

Codebase-wide bug hunts and security audits, large migrations and modernization (framework swaps, API deprecations, language ports spanning thousands of files), and dead-code cleanup and refactoring — work that was hard to do in a single pass now fans out in parallel.
How to start

Ask Claude directly (for example, “Create a workflow”), or enable the ultracode setting from the effort menu (it sets effort to xhigh) and let Claude decide when to use workflows.

Notes

Research preview — rolled out May 28, 2026.
Where it works — available in the Claude Code CLI, Desktop, and VS Code extension, and on the Claude API, Amazon Bedrock, Vertex AI, and Microsoft Foundry.
Supported plans — Max, Team, and Enterprise. Workflows are disabled by default on Enterprise; an admin must opt in via settings.
Token usage — dynamic workflows can consume substantially more tokens than a typical Claude Code session, so start on a scoped task to get a feel for usage in your work.

§ 4

Frequently Asked Questions

frequently asked

§ 4.1

What are dynamic workflows in Claude Code?

Claude breaks a problem into subtasks and fans them out across tens to hundreds of subagents running in parallel, dynamically writing orchestration scripts in a single session and checking its work before anything reaches you.

§ 4.2

How do I start a workflow?

Ask Claude directly (for example, "Create a workflow"), or enable the `ultracode` setting from the effort menu (it sets effort to xhigh) and let Claude decide when to use workflows.

§ 4.3

Which plans support dynamic workflows?

Max, Team, and Enterprise plans. On Enterprise it is disabled by default and an admin must enable it in settings.

§ 4.4

Where does it work?

In the Claude Code CLI, Desktop, and VS Code extension, and on the Claude API, Amazon Bedrock, Vertex AI, and Microsoft Foundry.

§ 4.5

What is it good for?

Broad, verification-heavy work like codebase-wide bug hunts and security audits, large migrations spanning thousands of files, and dead-code cleanup and refactoring.

§ 4.6

Anything to watch out for?

It can consume substantially more tokens than a typical Claude Code session, so start on a scoped task to get a feel for usage. It is currently a research preview.