Run, Approve, and Kill: How Bryn's Control Plane Keeps You in Charge

Before Run mode, Approve mode, or the kill switch mean anything, it helps to know exactly what they're controlling. The docs break a Play into three parts:

Plays exist in the first place because knowing about a good-fit account isn't the same as doing anything about it:

Most Plays live here. Once you've approved a Play, in Run mode it fires end to end the moment its trigger and conditions match, posting the Slack alert, updating the CRM record, sending the outreach draft, without waiting on a click from anyone. That's the entire point of setting the Play up once: the work of noticing and deciding is already done, so acting on it doesn't need to wait for a person to be at their desk.

Run mode isn't unbounded, though: it only ever runs inside the trigger and conditions you set. A Play with narrow conditions simply never fires outside them, in Run mode or otherwise.

Some actions are worth a second set of eyes before they leave the building: outreach to a prospect, a write to a system of record you care about. Approve mode is the opt-in setting for exactly that case. Bryn still does everything up to the action itself: it watches, scores, and drafts. It just holds there instead of sending, per the anatomy quote above, "outbound actions can hold for a human review first", until you give it a one-click yes.

You choose Approve mode per Play, not globally, so a low-risk internal notification can stay in Run mode while a customer-facing outreach Play requires your review every time.

Underneath both modes sits one override that beats them both, instantly:

You can stop everything at once, or scope the stop to one region, one segment, or a single Play, whatever the situation actually calls for. And the stop itself isn't silent: it's written to the same record everything else is, so there's never a question of whether something was paused and when.

Every action a Play takes, in either mode, is written to a permanent record: what the Play saw, why it scored the way it did, and what it actually did about it.

Two specific mechanics worth knowing, because they're the kind of detail that matters the first time something goes slightly wrong. First, writes to your CRM are additive, not destructive:

Second, the score behind every action is never hidden from you, and it never leaves your workspace either:

In plain terms: what your site or your team ever sees is the account and the named signals, Bryn's word for a plain, named read on a behavior pattern, like a repeat pricing visit, that led to a decision. The raw composite number stays internal to the audit record, for anyone who wants to check the reasoning later.