Review of LocalRAG
LocalRAG is a mobile app (iPhone, iPad, Android) that lets you ask questions to your documents (PDF, Word, Excel, EPUB, images, 15 formats) using a local LLM (Qwen3 4B) running entirely on the device. Users keep full privacy: no data leaves the device. For higher quality, the app can switch to the Claude API. Citations, PDF highlights and cross-language search are built in.
LocalRAG: Discutez avec vos documents en local, sans cloud
Best for
- Professionals mobile and often offline
- Lawyers reviewing confidential contracts
- Researchers reading many papers
- Students with long textbooks
- Anyone valuing privacy
Not ideal for
- Users wanting Claude quality without a key
- Desktop-only workflows
- Very long documents in pure local mode
- Use cases requiring cloud APIs/integration
Pros & cons
- ✅ Local LLM on-device (Qwen3 4B), fully offline
- ✅ Supports 15 formats (PDF, Word, Excel, EPUB)
- ✅ Citations and PDF highlight built in
- ✅ OCR for photos and scans
- ✅ Multilingual search via E5
- ✅ Optional Claude API fallback
- ⚠️ Local model limited to 4B parameters
- ⚠️ 3 GB LLM download required
- ⚠️ Slower than cloud models
- ⚠️ No native desktop version
- ⚠️ Claude API requires a paid key
Our verdict
LocalRAG is one of the most convincing solutions on the market for confidential mobile document research. The ability to run a local LLM on iPhone, iPad or Android with no cloud calls is a major asset for lawyers, researchers, doctors and any professional handling sensitive documents. Compatibility with 15 formats, built-in OCR, sourced citations with PDF highlight and multilingual search make it a comprehensive tool. The 4B local model quality remains below cloud models like Claude, but the app allows switching when needed. For anyone seeking a private, mobile and functional alternative to classic document chat tools, LocalRAG is now a serious reference worth installing absolutely today.
Alternatives to LocalRAG
- Productivity suite with built-in AI: summaries, writing, turning notes into tasks, workspace search, and faster execution for teams.Editor’s pickProject Management+3
- Cloud business phone system with built-in AI for sales and support teams.AI AssistantSMBs+1
- AI-powered PR assistant that automates pitches to journalists to land hands-free coverage in top publications.Lead Generation+3
- OpenAI's frontier model combining reasoning, coding and native computer-use with a 1M-token context.AI AssistantAI Agents+2
- Enterprise AI agent platform to automate workflows and business operations at scale.AI Agents+2
- Free AI companion platform with natural chat, image and video generation for adults seeking a personalized experience.AI Assistant+3
- All-in-one AI photo studio for headshots, photo enhancement, signatures and creative design.Image Generation+2
- AI presentation generator that builds dynamic, immersive slides in seconds.Content Creation+2
- Realistic AI companion chat platform offering personalized characters and immersive 24/7 conversations for adult users.AI AssistantStorytelling+1
- Realistic AI companion with customizable characters and immersive 24/7 chats for adult users.AI AssistantStorytelling+1
- AI companion focused on the furry community, offering anthro and feral characters for ongoing personalized chats.AI AssistantStorytelling+1
- AI PR assistant that lands DR30+ backlinks in Forbes, Business Insider, and TechCrunch through automated journalist outreach.Content Strategy+3
Read also
FAQ
Does LocalRAG really work offline?
Yes, the local Qwen3 4B LLM runs entirely on your device after a one-time ~3GB download. No data is sent to the cloud.
How many document formats are supported?
LocalRAG supports 15 formats, including PDF, EPUB, Word, Excel, PowerPoint, images and scanned documents thanks to built-in OCR.
Can I use Claude for higher quality?
Yes, you can switch to the Claude API (Opus, Sonnet, Haiku) using your own key. Only relevant context is sent.
Is there a free trial?
Yes, LocalRAG includes a one-week free trial that unlocks all features on iOS and Android.
Do my documents leave my device?
No, documents are processed locally for extraction, indexing and search. With Claude API, only relevant portions are sent.