Your agent has been exploring, editing, running tests — and now you’re 40 messages in and the answers are getting vague. Context rot has set in.
Subagents fix this. A subagent is a fresh agent that the main agent spawns for a specific task. It gets its own context window — clean, focused, no history from the parent session. When it’s done, only the result comes back. Your main context stays light.
Two reasons this is a game-changer: — Context doesn’t fill up. A subagent exploring 15 files doesn’t pollute the parent’s context with all 15. The parent only sees the summary. — Parallelism. You can run multiple subagents at once — one exploring the backend, one checking the frontend, one running tests. They work simultaneously.
Every major coding agent now supports this: Copilot CLI, Claude Code, Codex, Cursor, OpenCode. The pattern has converged. In Copilot CLI, it’s the task tool. In Claude Code, the Task tool with mode: "background".
If you’ve been following tips #16 and #17 — subagents are the automated version of both.
💡 Try asking for a subagent next time you have a concrete subtask: “use a subagent to find all the places we handle authentication errors.” 🔗 VS Code Copilot subagents docs
Have you used subagents / the task tool?
🟢 Yes — I delegate regularly
🟡 Tried it once or twice
🔴 Didn’t know my agent could do this
⚪ I prefer doing everything in one session
What would you delegate to a subagent?