Dynamic Workflows in Claude Code: How Anthropic Just Made Months-Long Engineering Work Finish in Days
Claude Code can now spin up hundreds of parallel agents on a single task — and the proof is a 750,000-line Bun rewrite that took 11 days.
In this article
Updated June 4, 2026: This article originally discussed these features while in preview. As of late May 2026, dynamic workflows and Opus 4.8 upgrades are now in general availability. Content below has been updated to reflect GA status.
Anthropic just shipped dynamic workflows in Claude Code — and the headline example says everything you need to know. Jarred Sumner used it to port Bun from Zig to Rust: 750,000 lines of Rust, 99.8% of the existing test suite passing, eleven days from first commit to merge. Work that would have taken a team of engineers a quarter. Done by Claude, in under two weeks.
What Dynamic Workflows Actually Are
The core idea: some problems are too big for a single agent pass. A bug hunt across an entire service. A migration touching hundreds of files. A plan you want stress-tested from every angle. Dynamic workflows handle these by having Claude dynamically write orchestration scripts, break the work into subtasks, and fan them out across tens to hundreds of parallel subagents — all in a single session.
Each subagent works independently, and other agents actively try to refute what they find. Results are verified before they're folded in. You don't see partial results or intermediate noise — you come back to a single, coordinated answer. And because progress is saved as the run goes, a job that's interrupted picks up where it left off. No starting over on a 10-hour migration because your laptop went to sleep.
The Bun Rewrite: What a Real Dynamic Workflow Looks Like
Jarred Sumner's Bun port from Zig to Rust is the clearest proof of concept anyone's published. Here's roughly how the workflows chained together:
Step 1: One workflow mapped the right Rust lifetime for every struct field in the entire Zig codebase. Step 2: The next wrote every .rs file as a behavior-identical port of its .zig counterpart — hundreds of agents working in parallel, two reviewers on each file. Step 3: A fix loop drove the build and test suite until both ran clean. Step 4: An overnight workflow addressed unnecessary data copies and opened a PR for each.
That's not Claude helping a developer write code. That's Claude doing the engineering work while the developer reviews the output. The distinction matters.
What You Can Use It For Today
Anthropic calls out three main buckets:
Codebase-wide audits: Bug hunts, security hardening, profiler-guided optimization, dead code identification. Claude searches in parallel and runs independent verification on every finding — so the report surfaces real issues, not false positives from a single-pass scan.
Large migrations and modernization: Framework swaps, API deprecations, language ports spanning thousands of files. This is where dynamic workflows fundamentally change the economics of legacy codebase work.
High-stakes verification: When you need a wrong answer to be caught before it reaches you — not after — workflows give Claude independent attempts at the same problem and adversarial agents working to break the result. Think: infrastructure changes, financial logic, compliance-sensitive code.
Ultracode mode: The easiest way to trigger dynamic workflows is to turn on ultracode in the Claude Code effort menu. It sets effort to xhigh and lets Claude decide automatically when a workflow is the right tool for the job. You can also just tell Claude directly: "Create a workflow."
Important: These Cost More Tokens
Anthropic is unusually direct about this: dynamic workflows consume meaningfully more usage than a typical Claude Code session. Hundreds of parallel agents add up. The recommendation is to start on a scoped task to understand your usage before throwing a full migration at it. In Claude Code, the first time a workflow triggers, it shows you what's about to run and asks for confirmation. Enterprise plan admins can also disable workflows org-wide from managed settings.
Availability
Dynamic workflows are in general availability today across Claude Code CLI, Desktop, and the VS Code extension. Available on Max, Team, and Enterprise plans (Enterprise is off by default — admins need to enable it), plus the Claude API, Amazon Bedrock, Vertex AI, and Microsoft Foundry. If you're on Max or Team, it's on by default. Just ask Claude to create a workflow, or flip on ultracode and let it decide.
The Bigger Picture
The gap dynamic workflows fills is real: between firing off a single subagent and building out a full agent team yourself. Historically, if you wanted parallel agents working a problem, you had to architect that orchestration layer yourself. Now Claude writes the orchestration. That shift — from using an AI tool to delegating an engineering project — is what makes this interesting.
Combined with Opus 4.8's improved judgment and honesty (4x less likely to let code flaws slide unchecked), dynamic workflows start to look like the first real version of autonomous engineering work. Not perfect, not unsupervised — but genuinely capable of handling the kind of work that used to require a team and a quarter.
Sources
Anthropic: Introducing Dynamic Workflows in Claude Code — claude.com/blog/introducing-dynamic-workflows-in-claude-code
Get the Claude playbook in your inbox.
One weekly email for Claude and Claude Code users. Real workflows, no hype. Subscribe and we send you The Claude Power-User Cheatsheet.
— ¶ —
Luke Thompson
Luke Thompson is the founder of The Operations Guide, LLC and editor of The Claude Insider. Based in Jonesborough, Tennessee, he has spent years building AI-augmented business systems and automation workflows for operators and teams. He began working with large language models in production well before the current wave of consumer AI tools, integrating them into client workflows, content pipelines, and operational infrastructure. At The Claude Insider, he writes about Claude with the specificity of someone who uses it daily as a professional tool — not as a reviewer or commentator, but as a builder. His coverage focuses on what actually works: prompt patterns, API integration strategies, agentic workflows, and the real-world tradeoffs that practitioners face. He is not affiliated with Anthropic, PBC.
Articles are researched and drafted with AI assistance, reviewed and edited by Luke Thompson.
Know where AI can pay off in your company.
Take the free two-minute AI Readiness Assessment. See your score, the two gaps holding you back, and the next move worth making.
