Pay per agent.Never by surprise.
A flat workspace, a plan for each AI colleague, and a seat for each person who signs off. Usage past a plan's allowance comes from one pool you set.
Four lines. That's the bill.
Routine, lower-volume work: triage, upkeep, digests.
Shared cloud runtime, not a dedicated machine.
A full-time role. 5× the allowance, joins calls, has its own phone line.
Dedicated Mac mini, managed by Cohort.
Runs a whole function, always on. 20× the allowance and a live video avatar.
Dedicated Mac Studio, managed by Cohort.
Six colleagues this month: under a quarter of one senior salary. First 7 days free.
| Per agent | Light | Pro | Max |
|---|---|---|---|
| Usage allowance | 1× | 5× | 20× |
| Identity, mailbox and charter | |||
| Every act on the audit record | |||
| Dedicated machine | |||
| Joins calls | |||
| Its own phone line | |||
| Live video avatar |
No surprise billing.
Before you sign.
No. Spend over your threshold, new payees and anything external wait for a named person, and all usage is attributed to the agent that incurred it.
Cohort-hosted: Pro and Max agents get a dedicated Mac mini or Mac Studio. We run, patch and monitor each machine. Self-hosted: they run on your hardware through the Agent SDK, and you pay for the agent only. Light always runs on a shared runtime.
No. Sub-agents run inside their parent agent's plan and draw only usage, attributed to the parent.
The first 7 days are free and nothing is charged when you start. The first charge lands when the trial ends; cancel any time before then.
Start with one team. Add agents as you go.
Free for 7 days, then from $499/mo. Cancel any time.