Skip to main content
Deal Brain already extracts observations and action items from everything it ingests. Signals let you go further and define your own extraction rules — patterns specific to how your team sells. A signal is a description of something you want to find. Deal Brain reads every eligible artifact and records an observation each time it finds a match.

​
What you’d use one for

Signals are useful whenever there’s a pattern you keep looking for by hand:
  • A competitor being evaluated alongside you
  • A specific technical requirement that changes scoping
  • Budget approval language that means a deal is real
  • A compliance or security requirement that triggers a different motion
  • A churn risk phrase you’ve learned to watch for
Once a signal is active, you stop hunting for these. They accumulate against your deals automatically.

​
How a signal works

1

You describe what to look for

In plain language: what counts, what doesn’t, and examples of both. This description is the whole configuration — there are no keyword lists or regex rules.
2

Deal Brain previews it

Before anything goes live, you can preview the signal against real artifacts to see what it would have caught.
3

You activate it

From that point on, every new eligible artifact is evaluated against the signal.
4

Optionally, you backfill

Run the signal across your existing history so you’re not starting from zero.

​
Key terms

​
Scope and permissions

Signals are organization-wide. A signal created by one admin applies to every artifact in the org, and observations are visible to everyone who can see the underlying deal. There are no personal signals. Creating and managing signals requires the configure_signals permission.

​
Where you work with signals

Two surfaces, acting on the same org-wide signals:
  • The Deal Brain app — open Signals in the left sidebar. Manage Signals is where you describe a new signal in plain language and see what’s configured. Collected Signals is where you read the observations your signals have recorded, with the source quote behind each one.
  • Your AI platform over MCP — where an agent drafts a signal, previews it against artifacts you name, activates it, and backfills your history.

​
Where to go next

Create your first signal

Both paths: describe a signal in the app, or draft, preview, and activate one over MCP.

Write good instructions

The difference between a signal that works and one that returns noise.

Lifecycle and backfill

Statuses, transitions, and running a signal over your history.

Read what a signal found

Review observations and their source quotes on the Collected Signals tab.