The Problem No One Talks About
Think about the last time you uploaded a document to an AI chatbot. Maybe it was a contract you needed summarized, a research paper you wanted to query, or meeting notes you needed to extract action items from.
You got your answer in seconds. It felt like magic.
But here's what happened behind the curtain: your document — that contract with confidential terms, that proprietary research, those internal meeting notes — was transmitted to a server you don't control, processed by a model you can't audit, and potentially retained for purposes you didn't consent to.
Every document you upload to a cloud AI becomes a data point in someone else's business model.
This isn't a hypothetical concern. It's the fundamental business model of "free" AI. You're not the customer — you're the product. Your documents are training data. Your questions reveal your interests. Your workflows become features in their next release.
The Subscription Trap
And then there's the cost. ChatGPT Plus: $240/year. Notion AI: $120/year. Copilot for Microsoft 365: $360/year. These aren't tools you own — they're toll booths on your own productivity.
Miss a payment? Lose access to your own workflows. Change providers? Start from scratch. Want to use it offline? Not possible.
We've normalized a model where you pay monthly to access your own intelligence. That should feel wrong, because it is.
What "Local-First" Really Means
Local-first isn't just a technical architecture — it's a philosophy. It means:
- Your data stays on your machine. Not "encrypted in the cloud." Not "we promise we don't look." On your hard drive, in a SQLite database you can inspect.
- You own the software. One payment. No subscription. No "Basic/Pro/Enterprise" tier where the features you need are always in the next tier up.
- It works without permission. No login required. No account creation. No "verify your email." Download, install, use.
- You choose the AI. BYOK — Bring Your Own Key. You pick the model. You control the connection. You set the budget.
Why We Built Recall.OS
Recall.OS started from a simple frustration: we had hundreds of PDFs, research papers, video lectures, and notes scattered across our machines. Finding specific information meant opening files one by one, Ctrl+F-ing through documents, or hoping we remembered which file had that one paragraph.
The AI-powered alternatives all required uploading everything to a cloud service. For personal notes, maybe that's acceptable. For client contracts, proprietary research, medical records, legal documents? Absolutely not.
So we built what we wanted:
A desktop app that indexes your documents locally, lets you ask questions in plain English, and gives you answers with exact page-number citations — without any of your data ever leaving your machine.
The Architecture of Privacy
Privacy in Recall.OS isn't a feature toggle. It's structural. Here's how:
SQLite for everything. Your document index, embeddings, and conversation history live in a single SQLite database on your device. No cloud database. No sync service. Just a file on your disk that you could back up with a USB drive.
Hybrid search. We combine vector similarity search (using sqlite-vec) with full-text search (FTS5) using reciprocal rank fusion. This gives you better results than either approach alone, and it all runs locally.
Tauri, not Electron. Recall.OS is built with Tauri v2 and Rust. The result: a 10MB installer instead of 200MB. 50MB of RAM instead of 500MB. Native performance instead of a Chrome tab pretending to be an app.
BYOK. The only external connection is to the Gemini API — and it uses your API key, sent directly from your device to Google. We're not a middleman. We don't see your queries. We can't access your key. Google's free tier is generous enough for most personal use, so your "AI subscription" cost is effectively $0.
Who This Is For
We built Recall.OS for knowledge workers who are drowning in documents:
- Researchers managing hundreds of papers and needing to find that one methodology detail from a paper they read six months ago.
- Law students and legal professionals who need to query case law and contracts without exposing confidential information.
- Analysts synthesizing information across reports, presentations, and data sources.
- Anyone who has ever spent 20 minutes looking for a specific passage they know exists somewhere in their files.
If you manage more than 50 documents and you care about where your data goes, Recall.OS was built for you.
The Sovereign AI Movement
Recall.OS is our first product, but the vision is bigger. We believe we're at an inflection point: the AI industry is consolidating around cloud-first, subscription-locked, data-harvesting models. But there's a growing counter-movement of people who want AI tools that work for them, not on them.
We call this sovereign AI: software that respects your autonomy, runs on your hardware, and answers only to you.
Project Intuitus is building a suite of these tools. Recall.OS is the beginning — your personal AI memory. More tools are coming, each designed with the same principles: local-first, sovereign, BYOK.
Try It
Recall.OS is available now for Windows. The free trial includes all features with up to 25 documents. A one-time $29.99 license unlocks unlimited documents forever.
No account required. No credit card for the trial. No tracking. Just download and start indexing.
Ready to own your AI memory?
Download Recall.OS and start searching your documents locally.
Try Recall.OS FreeIf this resonated with you, share it with someone who cares about privacy. The sovereign AI movement grows one person at a time.