A local-first intelligence system that reads your body of work, maps your intellectual architecture, and surfaces what you need — before you ask. Runs entirely on your machine. Nothing leaves.
Knowledge workers accumulate files the way cities accumulate infrastructure — constantly expanding, rarely indexed, almost never understood as a whole. After years of work, the files are there. The intelligence locked inside them isn't accessible.
Spotlight finds the file. Notion organizes it. Obsidian links it. None of them know that the document from three years ago is the intellectual ancestor of the work happening today. None of them surface the pattern. None of them know what matters to you.
"The problem is not storage. The problem is surfacing. The intellectual architecture exists. It is simply invisible."
Three layers of extraction. Each one deeper than anything you've had before.
INDEX reads your files where they already live — no moves, no imports, no renames. Point it at a folder. It reads everything.
A local LLM runs on your machine, extracting summaries, entities, concepts, and intellectual threads from every document. Then it maps the connections between them.
INDEX infers your intellectual architecture automatically — your threads, projects, and frameworks — and scores everything against how you actually think. Not the average user.
Not a search engine. Not a file manager. Not a note-taking app. A cognitive prosthetic for your body of work.
Two modes, one database. Data Mode ranks by connection count. Intelligence Mode ranks by relevance to your inferred intellectual architecture — your threads, your projects, your entities.
The LLM runs locally via Ollama. The database is SQLite on your disk. No external servers, no API calls for extraction, no subscription required. You could disconnect from the internet and nothing would change.
Cross-file relationship mapping produces a graph that reflects the actual topology of your thinking — not just keyword co-occurrence. See the intellectual ancestor of your current work, three projects back.
INDEX reads files in place. No migration, no import, no renaming. Works with Pages, Word, PDF, Markdown, Keynote, PowerPoint, Excel, RTF, HTML — the formats you already have.
You wrote it. You should be able to use it. Your intelligence belongs to you — not to a platform, not to an AI company, not to a cloud.
No moves, no renames, no imports. INDEX reads what you already have, in the formats you already use.
No cloud, no subscription, no external servers. Nothing you own becomes someone else's training data.
Switch the underlying model tomorrow and lose nothing. The graph, the extractions, the architecture — all yours.
INDEX is in early access. Get in now — bring your body of work, we'll show you what's inside it.