Your meeting notes live in Granola. Your email lives in Gmail. Your account data lives in your CRM. None of them know what any of it means for your business. That's the job of a trace.
You or your agent decide what the Brain should know. Share a meeting note from Claude or ChatGPT, forward an email, tag @brainbox in a Slack thread — and BrainBox reads that event against the Brain and produces exactly one thing: the interpretation. Which accounts it touches. Which process it belongs to. What risk or signal it surfaces. What's worth remembering.
Nothing raw gets duplicated into yet another store, and nothing is captured behind your back. The transcript stays where it lives; the meaning goes into the Brain.
A trace always points back at the event that produced it. When the Brain says an account is at risk, you can see exactly which meeting — and which sentence — made it think so.
Tag @brainbox in Slack, or send from Claude or ChatGPT mid-conversation — every path lands the same way: interpreted against the Brain, written as a trace. No pipeline to build.
# sent to BrainBox: Tuesday sync note with Acme (from Granola)
→ trace written:
accounts: Acme Corp
process: renewal (Q3 cycle)
signal: budget approval moved to October — renewal timeline at risk
source: meeting 2026-07-07, attributed
Next week, any agent — or any teammate — asking about Acme gets the current picture, not a transcript search.
Grounded answers, not guesses — from a living model of your business.