Data

Your data stays with you.

Connect the files and apps you already use. Your node makes them useful on your machine. You choose what anyone else can see.

~/.topos/database.db

  1. Connect

    Files, plugins, AI history, and first-party connectors land as-is.

  2. Ingest

    Every record splits at the door: the full copy stays yours; a redacted copy is all anyone else can read.

  3. Refine

    On your hardware, records become people, dated facts, insights, and a timeline you can check.

  4. Answer

    Questions hit the refined layer — not a dump of exports — inside your boundaries.

Why it matters

Session tools forget. Your node does not.

Chat tools are sharp for a turn — then the context ends. A personal node keeps a database of what you connected, on disk, and reads it while you sleep.

  • ChatGPTa chat history you pasted in
  • Claudea conversation
  • Your nodewhat you connected — continuously

Point any of them at your node when you want a session that already knows you. The file stays yours either way.

Connect

Bring it in as it already exists

No format gate. No migration project. One source is enough to begin.

Six first-party sources today. Not “all your data.” Remote connectors ship when they ship.

How it works

Connect. Ingest. Refine. Answer.

What happens to a record after it lands on your node.

  1. Connect

    Files, plugins, AI history, and first-party connectors land as-is.

  2. Ingest

    Every record splits at the door: the full copy stays yours; a redacted copy is all anyone else can read.

  3. Refine

    On your hardware, records become people, dated facts, insights, and a timeline you can check.

  4. Answer

    Questions hit the refined layer — not a dump of exports — inside your boundaries.

incoming recordCognitiveFirewallyours · full copydinner w/ Maya, 7pm —call 555-0142 re: leasetheirs · redacteddinner w/ [person], 7pm —call [phone] re: lease

At ingest, every record splits: the full copy stays yours; a redacted copy is all anyone else can read.

The refinery

Raw in. Signal out.

Automatic sorting so what you keep — and what you share — is organized context, not a pile of exports.

  1. Normalize

    Every source becomes the same kind of record: who, what, when, where it came from.

  2. Resolve

    Mentions gather per person — “Mom”, “Linda”, and a phone number across apps lean toward one person your node knows.

  3. Distill

    Records are read for what is durable: facts with a source and a date. Compression is the product.

  4. Connect

    Facts stitch across time. Patterns surface as readings computed on your machine.

Your node’s data only. Never pooled. Never used to train on you.

Signal you can check

Receipts on every claim

Before you share it, you can verify it. Click through. Fix what’s wrong. Raw stays underneath.

  • Sourced. Every fact links back to the records it came from.
  • Dated. When it was true — not just that it was said.
  • Yours to correct. Wrong? Fix or remove it. Your call wins on every future read.

A fact with its receipts: sources, date, your edit control.

Behavioural readings are statistics from your own records — session rhythms, activity peaks — not conclusions about who you are.

What you control

You choose what leaves

Sharing is filtered and revocable between people who already have TOPOS. Someone asks, you approve a slice, you can stop it. There is no feed pushing your life outward.

  • Keys, not broadcasts

    Each person or AI you share with holds a key to a page — never the whole database.

  • A file on your disk

    Local-first SQLite. Unplug the network and the file is still readable.

  • Sources you can cut

    Disconnect any source anytime. What you connected stays under your control.

  • Open-source engine

    Inspect it. Self-host the parts that are yours to run.

How sharing stays narrow →

Start here

One install. Your file. Your rules.

Create an account, connect one source when you’re ready.

enter TOPOSSee how to connect →

Security · Pricing · Home