Day 25

Subagent Concepts

Learn when and why to spawn parallel AI agents instead of working sequentially.

0 / 5 tasks

What are subagents?

When a task has multiple independent parts, you can ask Claude to spawn parallel subagents — each working on a different part simultaneously. One agent writes the API, another writes the tests, another writes the docs. The results are merged at the end.

When to parallelize

Parallelize when: tasks are independent (output of A doesn't depend on B), when you have a set of similar tasks to process, or when you want independent verification (two agents review the same code for different issues).

Don't parallelize when: tasks are sequential by nature, when you have tight token budgets, or when the merge step is more complex than running sequentially.

Requesting parallel work

bash
claude
> I need to implement a new user profile feature. Split this into parallel tasks:
>
> Task A: Implement the backend API endpoint (POST /api/profile, PATCH /api/profile)
> Task B: Write the SQLAlchemy model and migration
> Task C: Generate pytest tests for the API endpoints
>
> These are independent — run them in parallel and show me the results of each.

Practice Tasks

0 / 5
← Module 4: MCP Servers ↑ Module overview Day 26: Running Parallel Tasks →