What does Anthropic add to Claude Code?

Anthropic adds “dynamic workflows” to Claude Code

Anthropic has released an update to Claude Code that introduces dynamic workflows, allowing the system to run hundreds of subagents in parallel for complex engineering tasks.

The feature is aimed at work that benefits from breaking a large goal into multiple coordinated steps—such as framework migrations, where code needs to be analyzed, transformed, tested, and validated across many components.

Why it matters

• More parallelism for software engineering: Instead of a single linear agent loop, dynamic workflows enable concurrent subtask execution, which can reduce end-to-end time for multi-stage engineering projects.
• Better handling of complexity: Migration work typically involves numerous edge cases; running multiple subagents in parallel is designed to improve coverage and coordination.
• More capable “agentic” tooling: Claude Code is part of a broader push toward developer-facing AI systems that can orchestrate work rather than just draft responses.

What’s included and what’s not

Anthropic’s update specifically highlights:

• parallel subagents
• dynamic orchestration behavior
• usage examples like framework migrations

The available details don’t specify the exact user controls (e.g., whether developers choose the number of subagents), performance metrics, or limits, such as how Anthropic handles failures or coordination conflicts.

Still, the move is notable because it turns Claude Code from a straightforward assistant into a more structured execution environment, bringing it closer to the multi-step workflow automation developers increasingly expect from agentic tools.