Anthropic launches Dynamic Workflows in Claude Code

Anthropic launched Dynamic Workflows in Claude Code on May 28, 2026, introducing a research-preview feature that orchestrates complex coding projects using hundreds of parallel AI agents working simultaneously within a single session. The new capability allows users to orchestrate large, multi-step coding projects through parallelized agents that operate on subtasks simultaneously. Dynamic workflows are available in the Claude Code command-line interface, desktop application, and Visual Studio Code extension for users on Max, Team, and some Enterprise plans, and Anthropic is also making the feature available through the Claude API and on Amazon Bedrock, Vertex AI, and Microsoft Foundry.

Claude AI code extension now available for Max, Team, and Enterprise plans with AI-powered coding assistance feature.

How Dynamic Workflows Changes the Architecture

The core architectural shift in Dynamic Workflows is a single design decision: the orchestration plan moves into executable code rather than living in the model’s working memory. A dynamic workflow is a JavaScript script that orchestrates subagents at scale, with Claude writing the script for the task you describe, and a runtime then executing it in the background. This approach contrasts sharply with previous implementations where intermediate results accumulated in Claude’s context window, limiting how many agents could coordinate before hitting capacity constraints.

When a workflow starts, Claude creates a plan from the user prompt and divides the job into smaller pieces, then assigns those tasks to parallel agents, while separate agents check or challenge the findings before the results are combined. Once Claude breaks a task into subtasks and dispatches agents to work in parallel, each approaching the problem from an independent angle, a separate layer of agents runs to refute those findings, with the run iterating until the results converge under that adversarial check before anything reaches the user.

Scale and Performance Demonstrated

The feature’s capabilities were demonstrated in a significant real-world application. Jarred Sumner used dynamic workflows to help port Bun from Zig to Rust, producing about 750,000 lines of Rust code, with 99.8% of the existing test suite passing, and completing the work in 11 days from first commit to merge. The rollout is aimed at software engineering work that is difficult to complete in a single pass, including codebase-wide bug investigations, security reviews, large migrations, and other jobs that may touch hundreds or thousands of files.

According to the official Claude Code documentation, a workflow can run up to 16 agents concurrently, with a hard ceiling of 1,000 total agents per run, with the official launch post describing runs spanning “tens to hundreds of parallel subagents,” consistent with those limits.

Access and Activation Settings

For Max and Team users, and for those using Claude Code through the API, the setting is on by default, while Enterprise customers need an administrator to enable it. Pro plan users can turn it on manually from the Dynamic Workflows row in /config, with Claude Code version 2.1.154 or later required on all platforms.

Users can trigger workflows in two ways. They can ask Claude to create a dynamic workflow directly (e.g., “Create a workflow”), or switch on a new Claude Code-specific setting called ultracode, which is accessible through the effort menu and sets the effort level to xhigh, while letting Claude decide automatically when to use a workflow to handle your task.

Use Cases and Early Applications

Early access users and teams inside Anthropic have been using dynamic workflows for a wide range of use cases, including codebase-wide bug hunts, profiler-guided optimization audits, and security audits where Claude searches a service or repo in parallel, then runs independent verification on every finding so the report surfaces real issues, with the same shape working for hardening passes including auth checks, input validation, and unsafe patterns across an entire codebase, as well as large migrations and modernization efforts where Claude can handle framework swaps, API deprecations, and language ports that span thousands of files end-to-end.

“Dynamic workflows” are designed for parallel work and long-running tasks that span hours to days, and can handle complex engineering tasks that would have previously taken weeks, with execution status saved periodically so interrupted work can be resumed from where it left off without having to start over.

Cost Considerations and Token Usage

Anthropic recommends starting with a limited task, noting that dynamic workflows can use substantially more tokens than a standard Claude Code session. If that model is Opus 4.8, at the same rate Anthropic kept flat from Opus 4.7, a 500-agent audit can shift the session bill by an order of magnitude compared to a standard Claude Code session. Both features consume meaningfully more tokens than a typical session, with a run able to spawn up to 1,000 agents, so costs climb fast.

The first time a workflow is triggered, Claude Code shows the user what is about to run and asks for confirmation, while organisation administrators can also disable workflows through managed settings.

Launch Alongside Claude Opus 4.8

Dynamic Workflows launched alongside Claude Opus 4.8, which Anthropic released just 41 days after the previous version. Anthropic released Opus 4.8, the newest version of its most advanced publicly available model, which is available everywhere, with standard pricing at the same level as the previous Opus release. Early testers said the model is “more likely to flag uncertainties about its work and less likely to make unsupported claims.”

Fast mode now supports Opus 4.8 at a lower price, with fast mode for Opus 4.8 three times cheaper according to Anthropic official release.

Key Facts

Launch date: May 28, 2026, as a research preview feature
Agent capacity: Up to 16 concurrent agents, with a maximum of 1,000 total agents per workflow run
Availability: Claude Code CLI, Desktop, VS Code extension, Claude API, Amazon Bedrock, Vertex AI, and Microsoft Foundry
Default activation: On by default for Max and Team plans; Enterprise requires admin enablement; Pro users activate manually
Real-world validation: Jarred Sumner ported Bun from Zig to Rust, generating 750,000 lines of code in 11 days with 99.8% test pass rate
Required version: Claude Code v2.1.154 or later
Pricing: Opus 4.8 standard pricing remains unchanged; Fast mode now 3x cheaper