The AI model itself lives on your device — not just "processed locally". Ask questions about your PDFs, Office docs, EPUBs, images, audio recordings and videos with no internet at all. Or use Claude when you want higher accuracy.
Available on iPhone, iPad, and Android
1-week free trial included
We fixed a bug where a failed import could still look like it succeeded: the document appeared in your list even though not a single character had been read, and questions about it went unanswered. We also fixed losing the copy you already had when re-importing a document under the same name failed. Switching to the on-device AI now starts downloading the model — previously it only changed the setting, and you found out it was missing when you asked a question. Search is fixed for languages that separate words with spaces, such as English and German. We also fixed two crashes affecting some devices (running the on-device AI, and saving documents), and replaced the built-in sample with something closer to a real contract.
Scanned documents are now recognised in your device's language. Recognition was previously fixed to Japanese and English, so documents in Korean, Chinese and other languages often came out unusable. Text files saved in encodings other than UTF-8 — Shift-JIS, EUC-KR, GB18030, Windows-1252 and more — now open correctly. We also fixed headings and tables coming out scrambled in PDFs that mix Latin and Asian scripts, added an automatic retry when the service is briefly busy, and made long scanned documents less likely to stall during import.
Import your documents, ask questions in plain language, and get AI-powered answers — all processed on your device.
"What are the termination clauses?" "Summarize chapter 3." Just type a question in plain language and LocalRAG! searches across all your documents to find the most relevant passages. No need to open files one by one — AI reads everything for you.
PDF, EPUB, Word, Excel, PowerPoint, images, scanned docs — 25 file formats in a single collection, searchable all at once. Whiteboard photos and scanned PDFs are automatically processed with OCR to extract text for AI-powered Q&A.
Every AI response includes page-level source citations like "Service-Agreement.pdf (p.12)" — so you always know exactly which document and page the answer came from. Never blindly trust AI — verify with the original source.
The built-in local LLM (Qwen 3.5 4B) runs entirely on your device. No internet connection needed to get an answer, and neither your documents nor your questions are ever sent to a server. Feel safe asking AI about confidential documents and personal files. Need higher accuracy? Switch to Claude API anytime.
Get key takeaways from long reports, manuals, and research papers in seconds.
Find termination clauses, non-competes, payment terms, and risks instantly.
Extract findings, understand methodology, and chat with academic papers.
One of the few AI tools with native EPUB support. Summarize chapters and ask questions.
Import photos of whiteboards, receipts, or documents. On-device OCR extracts text for AI-powered Q&A.
Import audio recordings of meetings. Ask "What was decided about the budget?" Get answers with timestamps to jump to.
Import lecture recordings (MP4, MOV). On-device transcription extracts the content. Ask the AI to explain concepts and find topics across hours of video.
Journalists, researchers, and lawyers can transcribe sensitive interviews entirely on-device. No upload, full source confidentiality.
PDF, Word, EPUB, whiteboard photos, scanned receipts — no matter the format, just add them to a collection. AI searches across all your documents at once.
Meeting PDFs, Word minutes, text memos — formats don't matter. Add them to a collection and ask "What was decided about that issue?" No more opening files one by one.
Import your unread PDFs and EPUBs together. Without reading a page, ask "What's the key point of this book?" or "Explain the method in Chapter 3." AI reads every page for you.
ISO PDFs, RFC text files, internal Word policies — load them all into one collection. Ask "What are the authentication requirements?" and LocalRAG! finds relevant sections across every document.
Insurance policies, user manuals, contracts, recipes, whiteboard photos, scanned receipts. Add anything to a collection. You may forget what's in there, but LocalRAG! never does. When you need it, just ask — it tells you which document has the answer.
Watch how LocalRAG! turns your documents into conversations.
Two ways to chat with your documents. Choose what fits your needs.
| On-Device AI | Claude API | |
|---|---|---|
| Internet required | No — fully offline | Yes |
| Document & question data sent to server | None | Question + relevant excerpts only |
| Answer quality | Good (4B model) | Excellent (Opus / Sonnet / Haiku) |
| Response speed | ~10 sec to a few min | ~3 sec |
| Model download | ~3 GB (one-time) | Not required |
| Platform | iOS & Android (beta) | iOS & Android |
Both modes process documents (text extraction, indexing, search) entirely on your device. The on-device AI takes this further by generating answers locally too — so your documents, your questions, and the answers all stay on your device. Separately from either mode, the app sends usage statistics — tied to a random per-install ID, never to you, and never including your content; our Privacy Policy lists exactly what those contain.
"I imported 12 textbook PDFs and asked 'Explain the difference between TCP and UDP.' Got a clear answer with page numbers — no more flipping through 3,000 pages."
"I loaded 30 papers into one collection and asked 'Which studies use transformer-based models for medical imaging?' Found 8 relevant papers in seconds."
"I reviewed a 120-page NDA by asking 'Are there any non-compete clauses?' It found two — on pages 34 and 78 — that I would have missed in a manual review."
"Our team's SOPs, compliance docs, and meeting notes are all in one collection. New hires just ask 'What's the approval process for vendor contracts?' and get the answer instantly."
Powerful document AI that respects your privacy.
Documents (PDF, EPUB, DOCX, XLSX, PPTX, TXT, MD, CSV, RTF, HTML), images (JPG, PNG, HEIC, WebP), audio (MP3, M4A, WAV, AAC, FLAC) and video (MP4, MOV, M4V) — plus scanned and copy-protected PDFs.
Ask questions about your documents in plain language. Get answers with source citations.
AI answers show [1][2] source references you can tap. Jump directly to the cited page with yellow highlight on the exact passage.
All document processing happens on your device. Your files never leave your phone or tablet.
Okapi BM25 search with on-device embedding model (E5 Multilingual) for semantic search across languages. Ask in Japanese, find answers in English documents.
Import photos of whiteboards, receipts, or scanned documents. On-device OCR extracts text — and you can edit the extracted text before indexing.
Organize documents in nested folders up to 3 levels. Parent scope includes all child documents for focused Q&A.
Built-in local LLM runs entirely on your device — no internet needed, no API calls, and neither your documents nor your questions are ever transmitted. True zero-cloud document AI.
Use the on-device AI for fully offline operation, or choose Claude Sonnet 5 / Haiku 4.5 / Opus 5 for higher accuracy.
Three simple steps to chat with your documents.
Add documents from your file manager or other apps. 25 formats supported.
LocalRAG! extracts text and builds a search index — entirely on your device.
Ask anything about your documents. Get AI-powered answers with citations.
All the document types you work with, in one app.
Document processing — text extraction, chunking, and search indexing — happens entirely on your device. Your files are never uploaded anywhere. When you use the on-device AI (local LLM), answers are generated on your device too — no document text, no question text, and no answer text is ever transmitted. When you use Claude, only your question and the top 10 most relevant text snippets (~500 characters each) are sent to the Claude API. The full document is never transmitted.
Cloud-based services require you to upload your entire document to their servers. LocalRAG! keeps your files on your device and only sends small, relevant excerpts when you ask a question. Plus, with the built-in on-device AI, answers are generated locally too — your documents and your questions never leave your device. This makes it suitable for sensitive documents like contracts, medical records, or confidential research.
No. Anthropic's API usage policy explicitly states that API inputs and outputs are not used for model training. This is different from the free consumer Claude chat product. Your document excerpts are processed and discarded.
Yes. The free plan includes 5 questions per day with Claude Haiku and the on-device AI. You can also bring your own Anthropic API key (BYOK) for unlimited access to all models at no subscription cost — you only pay Anthropic directly for API usage.
LocalRAG! includes a built-in on-device language model that runs entirely on your iPhone, iPad, or Android device (beta). Once you download the model (~3 GB), you can ask questions about your documents with zero internet connection. Answers are generated on your device — neither your documents nor your questions are transmitted. It's available on all plans.
LocalRAG! supports 25 formats across four categories: 11 document formats (PDF, TXT, MD, CSV, RTF, HTML, HTM, EPUB, DOCX, XLSX, PPTX), 6 image formats (JPG, JPEG, PNG, HEIC, HEIF, WebP) processed with on-device OCR, 5 audio formats (MP3, M4A, WAV, AAC, FLAC), and 3 video formats (MP4, MOV, M4V) transcribed on-device with WhisperKit (iOS) and whisper.cpp (Android). Scanned and copy-protected PDFs are also supported.
Start free with 5 questions/day, or subscribe for more.
Free, Standard ($4.99/mo), and Pro ($9.99/mo) plans available. Cancel anytime.
Terms of Use · Privacy Policy