A team that runs itself — one objective, many agents.
A Playbook runs a whole objective end to end, routing each step to the right agent under your autonomy controls. Research, qualify, reach out, book, update the CRM — coordinated across your agents, not stitched together by hand.
A single agent does one job. A Playbook coordinates several — handing each step to the agent best suited to it, holding for your approval where you want a human in the loop, and advancing on its own where you don't. You define the play once; the agents run it.
- 01Compose
- 02Control
- 03Adapt
- 04Prove
Compose · Multi-agent steps
A Playbook is an ordered set of steps, each with a capability (research, outreach, booking, CRM update) and the agent role that owns it. Start from a built-in Playbook — Sales, Marketing, Customer Success, Front Desk — or build your own from scratch.
- 1.1Built-in Playbooks for sales, marketing, service & front desk
- 1.2Build your own: add, reorder & route steps to any agent
- 1.3Each step routed to the right agent automatically
Control · Autonomy you dial
Every step honors your autonomy controls. Set a step to act automatically, or hold it for human approval — review steps queue in one approval inbox and the run waits for your go-ahead. Nothing acts beyond the level you set.
- 2.1Per-step autonomy: auto, review, or off
- 2.2Review steps hold the run for human approval
- 2.3One unified inbox for every approval, any source
Adapt · Branching & recovery
Playbooks aren't rigid. Branch on outcomes — a positive reply books a meeting; a negative one routes to nurture. When a step fails, choose what happens: retry, skip, escalate to a human, or pause. The run adapts instead of stalling.
- 3.1Conditional branching on step outcomes
- 3.2Per-step failure handling: retry · skip · escalate · pause
Prove · Test before live
Try a Playbook in shadow before it acts: a guided activation walks you through readiness, runs it on one contact, and shows exactly what each agent would do — sending nothing — so you go live with confidence.
- 4.1Shadow runs: see what would happen, send nothing
- 4.2Guided activation: readiness → test → review → live
- 4.3Run performance: completion, stalls & timing per Playbook
Frequently asked
Set the play. Let the team run it.
One objective, many agents, under your controls. Test it in shadow, then go live.