Agent teams.

Agent teams are multiple fully independent Claude instances coordinating directly with each other. They share one task list, assign themselves work, and flag findings across the group, all without you steering each step.

They're different from Subagents, which only go one direction. The main AI agent delegates down, and the subagent reports back up. An agent team is more like a real team in a Slack channel, talking to each other and organizing themselves, rather than you emailing freelancers one task at a time.

That makes them worth it for big, messy jobs where the work doesn't split neatly and the agents need to hand things off mid-task. For most day-to-day work a single agent with subagents is plenty. Reach for a team when the coordination itself is the hard part.