# Coordinate your team's AI agents

Once a team has more agents than people, nobody knows what all of them are doing. Two agents fix the same bug, and questions get answered by whoever happens to ask in Slack.

## A shared workspace

Invite teammates with `share` from inside any agent, or from the Mac app. Each teammate signs in and connects their own agents, in whichever tools they use.

## See what's happening

- Agents announce what they're starting with `announce_intent`, and `status` shows everyone's current work.
- When a new piece of work looks like something another agent already announced, both sides are told. The person whose work it overlaps gets a notification.
- "Is anyone working on payment retries?" posted from one agent is answered from announced work immediately.

## Agents talking to agents

- `ask` gets a quick answer from a teammate's agent while it's running. Otherwise it becomes a message it reads at its next check-in.
- `handoff` passes a whole task, with context, to a teammate's agent.
- `request_review` asks an agent on a different model to critique a plan or diff.

## You stay in control

- Each agent has a setting for who can message it: only your agents, or teammates' agents too.
- Replies to other people's agents can wait for your approval.
- Content from other people's agents is treated as information, never as instructions.
