Subagents - Claude Code Blog

Claude Code has six ways to extend it — CLAUDE.md, skills, subagents, MCP servers, hooks, and plugins — and they overlap just enough to be confusing. This decision guide explains what each one is, when to reach for it, when not to, and how to combine them.

Claude Opus 4.8 pushes the autonomous horizon from an hour to a workday, tightens subagent coordination, and ships meaningful gains on cost-per-task. Here's what's actually different, what to do with it, and where it still falls short.

Plan Mode is the difference between agents that ship the right thing and agents that confidently ship the wrong thing. Here's how it works, when to use it, and the habits that turn it into your biggest reliability lever.

Worktrees are the single biggest unlock for running multiple Claude Code agents at once without conflicts. Here's how they work, when to use them, and the failure modes that take down most teams the first week.

Ultrareview (/ultrareview) runs a multi-agent cloud review of your current branch or a GitHub PR — parallel specialists reviewing security, architecture, correctness, and style. Here's what it is, how to run it, and when to reach for it.

Master Claude Code's subagent system to orchestrate specialized AI teams. Learn how to parallelize work, chain agent outputs, and build multi-agent workflows that tackle complex development tasks autonomously.