Agent collaboration
The agents are not four separate tools — they are one team. Work flows between them the same way it would between colleagues, without you orchestrating it.
The shared mailbox
Agents communicate through a shared mailbox. One agent sends another a message — “closed Acme, they signed for €4,000, kickoff next week” — and that message waits in the recipient's inbox until it next runs.
Every agent has three mailbox actions available to it:
send_message_to_colleagueHand a piece of work, with context, to another agent.check_messagesRead its own inbox.mark_message_actedAcknowledge a message and note what was done about it.When an agent starts a task, any high-priority unread messages are pulled straight into its working context — so an urgent handoff is acted on immediately, not whenever someone happens to check.
Where handoffs are already wired
Some handoffs are too important to leave to chance, so they are built in — they fire automatically when the triggering event happens:
An incoming reply is classified as interested, a price inquiry or a meeting request.
A lead is marked converted — the new client needs onboarding, documents and invoicing.
An invoice slips more than 30 days overdue — Sales owns the relationship and the chase.
A campaign finishes — the Project Manager runs the retro and extracts what worked.
Chaining agents into workflows
Built-in handoffs cover the common cases. For anything specific to how you work, the automation builder lets you draw the workflow yourself on a visual canvas — each node an agent action, the arrows the flow between them.
A typical chain: when a lead is qualified → Sales drafts a reply → wait 24 hours → if still no reply → Marketing sends a follow-up. You draw it once; the agents run it every time.