How do you lead a larger change with several coding agents?
Answer out loud or in your head first, then compare.
Answer
First I define the goal, the scope and a shared data format. I split the work where the parts are truly independent, e.g. the interface and the backend once the contract is agreed. Each agent gets a limited scope and criteria, and shared files have one owner.
Separate branches or worktrees protect against accidental overwrites, but they don't solve whether things work together. After merging, I run the shared tests and check the whole user journey. The review also covers changes to tests and migrations.
If the task is small, or all the steps depend heavily on each other, one agent may be better. I don't judge a solution by the number of workers, but by quality, time, cost and the effort needed to check it.
Why it matters: Keeps an advanced scenario without starting from orchestration tools.
Possible follow-ups: What do you do when the backend needs to change the contract halfway through?
How did it go? Be honest — “Not yet” puts the question on your review list.
Start with
- D-P15 · What is idempotency? — Backend, data and system design
- G-P01 · What is RAG? — RAG and retrieval
- H-P01 · How does an AI agent differ from a plain model call? — Agents and workflows
Further reading
- Anthropic: Effective harnesses for long-running agents(opens in a new tab)26.11.2025. Working in stages, progress artifacts and resuming between sessions.
- Claude Code: Create custom subagents(opens in a new tab)Docs; accessed 22.10.2026. Separate context, tools and permissions; details depend on the mode.