Every AI assistant will need your memory.
You should own it.
Trace turns your real conversations — messages — into a private memory base, one living profile per person: who they are, what you promised, what matters. On your device. Usable by any AI you choose.
Anna Kovalenko
- Partner at an early-stage fund; ex-founder (logistics)
- Fund thesis: infrastructure for AI agents
- Expects the one-pager tomorrow morning
Today Saturday, 30 August
- Decided — go ahead with the design-partner cohort in September
- Promised — one-pager to Anna tomorrow morning
- Paid — €40, co-working day pass
- Met — 7 conversations across 3 channels
Tasks 7 open · all auto-created from conversations
Projects people and promises, grouped by what you're working on
Connectors one thin adapter per channel
Problem
Your relationships live in chat scroll. You can't search your own life.
Professionals get a CRM. You don't.
A salesperson with 50 accounts gets pipelines and reminders. You, with 500 relationships, get an unsearchable scroll.
You met 30 people last month. You remember four.
Somewhere in those chats is a promise you made three weeks ago — you'll find out when they ask why you didn't keep it.
Personal CRMs all died the same death.
Nobody logs their own life by hand. Manual entry killed every one of them.
AI has none of your context.
Uploading your whole life to someone's cloud is exactly what people refuse to do — and they're right to refuse.
Solution
A CRM for your whole life that builds itself.
Trace captures every conversation and files it automatically, organized per person — stored privately on your device, zero data entry, ever.
AI turns raw history into living profiles: who this person is, what you promised each other, who they know, when their birthday is. Every night, a review distills the day — decided, promised, paid — and open promises become tasks on their own. Nothing falls through.
Day to day it feels like a personal CRM with no admin. Structurally, it's something bigger: a context layer any AI assistant can read — with your permission, on your side of the table. Models are becoming a commodity. Your memory isn't.
How it works
Five steps, one loop, zero effort.
Connect your channels Telegram live
Real-time capture plus full history backfill. Calls and meetings arrive as transcripts, summarized.
Conversations file themselves
One living profile per person — every message, call, and meeting on one timeline. No forms, no tagging, no data entry.
AI turns history into memory
Who they are, what to remember, what you promised — birthdays and contacts extracted, connections mapped across your network.
The loop closes every night
A daily review distills what was decided, promised, and paid — and every open commitment becomes a task automatically.
Any AI can plug in
An open memory API lets the assistant you already use read your context — with your permission, revocable any time.
Connectors
Every channel you talk in. One memory.
The core is source-agnostic — each new channel is a thin adapter, not a rewrite. Telegram proved the loop; the rest is breadth.
Roadmap order set with design partners. One connector ≈ one adapter — the hard part is already built.
Private by design
Your memory. Your device. Nobody else's business.
Lives on your device.
No central database of your private life to breach, sell, or subpoena. Capture and enrichment run for you, not on you.
We never monetize your data.
We can't — we don't have it. You pay for software and compute; your conversations never become the product.
Zero lock-in.
Full export any time, and an open memory API so the AI assistant you already use can read your context — with your permission, revocable.
Early access
Be one of the first hundred.
We're onboarding a design-partner cohort: the first hundred people who never drop a promise again. Telegram works today; WhatsApp, calls, and meetings are next.
Your conversations become your memory — on your device, owned by you. No forms, no tagging, no data entry. Ever.