A colleague whonever oversteps.
Put an assistant on a channel or a mailbox and it answers in your voice, records what the client said, and opens the deal. Everything it is not allowed to decide comes back to you as a request.
Free to start · no card · approved in 24h
Four limits, all of them switches
The first character you type ends its turn in that thread — for good, not for an hour.
Working days, working hours and your timezone. Out of hours it holds the reply rather than answering at 3am.
Moving a deal or writing to a field arrives as a request with the reasoning attached, and waits for a person.
A monthly spend limit per assistant. It stops and tells the owner rather than quietly running up a bill.
It will not commit youto anything.
Prices, deadlines and guarantees it was not given are blocked before sending — it offers to check with a colleague instead. Anger, a refund request or a legal question goes straight to a person with the context already summarised.
What it knows
Pulled out of real threads and attached to the contact or company as fields, with the message they came from.
Policies, prices and service rules, written once and quoted verbatim — split into client-safe and internal.
Distilled from deals your best closers actually won, and fed back into the first reply.
It reads only what the employee it acts for is allowed to read. Memory obeys the same scopes as the CRM.
Every run logsits own bill.
Tools called, tokens spent and the cost in cents — per conversation, in the activity log. Assistants run on a model key you connect or on the allowance included with the Scale plan.
Start with one channeland the limits on.
Watch a week of conversations, read what it wrote, then decide how much rope it gets.