Index your documents, videos, and screenshots. Ask questions in plain English. Get answers with exact citations — page numbers, timestamps, source links. 100% local.
PALM (Parallelized Attention-based Learning Model) is an approach that distributes transformer computations across multiple TPU pods, enabling training on models with 540 billion parameters. It achieves breakthrough performance on reasoning tasks through chain-of-thought prompting...
ML_Research.pdf — Page 47Vector search + full-text search combined with reciprocal rank fusion for superior accuracy.
Automatic transcription with timestamped citations. Jump to the exact moment.
Scanned PDFs and images become searchable text using Windows native OCR.
Every answer includes page numbers and source links. Click to jump directly.
Capture and index anything on your screen. Make your visual work searchable.
Your data never leaves your machine. No cloud uploads, no data harvesting.
Drop in PDFs, videos, audio, or set up watched folders for auto-ingestion.
Recall.OS extracts text, transcribes media, and creates searchable embeddings.
Query your knowledge base in natural language. Get instant answers.
Every answer includes citations. Click to jump directly to the original source.
Test before you commit
Unlimited, forever
Licensing is not yet available. Join the waitlist to get notified when it launches.
The free trial includes all features with up to 25 documents. Join the waitlist for license availability updates.
| Feature | Recall.OS | ChatGPT | Notion AI |
|---|---|---|---|
| 100% Local Processing | Yes | No | No |
| Your Data Stays Private | Yes | No | No |
| One-Time Payment | $29.99 | $240/yr | $120/yr |
| PDFs, Videos, Audio | Yes | PDFs only | PDFs only |
| Exact Page Citations | Yes | No | No |
| Works Offline | With API key | No | No |
| BYOK — No Vendor Lock-in | Yes | No | No |
Recall.OS uses Google's Gemini models for text embeddings, OCR, and generating answers. By bringing your own API key (BYOK), you control the connection directly — no middleman, no markup. The Gemini free tier is generous enough for most personal use. Get a free API key here.
Yes. All document processing, indexing, and search happens locally on your machine. Your documents are stored in a local SQLite database on your device. The only external connection is to the Gemini API for embeddings and generation — and that uses your own API key. Recall.OS collects zero telemetry, zero analytics, and zero user data.
Recall.OS is currently Windows-only. macOS support is planned — the app is built with Tauri v2, which supports cross-platform builds. Linux is also on the roadmap. Join the waitlist to be notified when new platforms launch.
Your existing documents remain fully searchable — you just can't ingest new ones past 25. Licensing is coming soon and will unlock unlimited documents with a one-time $29.99 payment. Join the waitlist to be notified when it's available.
License purchasing is coming soon. Once available, you'll receive a license key via email after purchase. Enter it in the app's License settings to unlock unlimited documents. Each license will support up to 5 device activations, with lifetime updates and a 30-day refund policy. Join the waitlist to be notified when licensing launches.
Download Recall.OS and start building your personal AI memory.
Requires Windows 10 version 1903 or later. Bring your own Gemini API key.