SafeRag SE is your personal AI assistant that runs entirely on your Mac. Built-in AI models, document intelligence, and encryption — no servers, no cloud, no setup. Just download from the Mac App Store and start chatting.
Requires macOS 15.0 or later • Built with SwiftUI • Apple Silicon only
Powered by llama.cpp running directly on your Mac. Download a model and start chatting — no external tools or server setup required. Your Mac's Neural Engine does all the work.
On macOS 26 and later, use Apple's on-device Foundation Models for fast, private AI responses. Seamlessly integrated — just select it as your AI provider.
Every conversation stays on your Mac. No cloud connections, no data transmission, no accounts. App Store sandboxed for additional security.
Upload PDFs, text files, and Markdown documents. Ask questions and get answers grounded in your own documents using RAG (Retrieval-Augmented Generation).
Protect your conversations with strong encryption. Set a passphrase and your chat history is encrypted at rest. Only you can read your messages.
Point SafeRag SE at a folder and new documents are automatically indexed. Build a personal knowledge base without manual uploads.
Get SafeRag SE from the Mac App Store. One click install, no configuration files, no Terminal commands.
Choose and download a model from within the app. On macOS 26+, Apple Foundation Models are available with no download needed.
That's it. Ask questions, upload documents, create prompt templates. Everything stays on your Mac.
Study with AI that understands your course materials. Upload lecture notes and papers, then ask questions. No data shared with cloud services.
Brainstorm ideas, draft content, and refine your writing with a private AI collaborator. Your creative work never leaves your machine.
Analyze documents, summarize reports, and prepare materials with AI assistance. Keep confidential work data on your Mac where it belongs.
Use AI without feeding your data to big tech. No cloud accounts, no usage tracking, no data mining. Just you and your Mac.
Download from the Mac App Store and start having private AI conversations in minutes. No setup, no accounts, no cloud.
Coming Soon on Mac App StoreFree with optional in-app purchases • macOS 15.0+
SafeRag SE includes a built-in AI engine powered by llama.cpp. You can download and run popular models like Llama, Mistral, Gemma, and more directly within the app. On macOS 26+, Apple Foundation Models are also available with no download required. You can optionally connect to Ollama if you have it installed.
Yes, SafeRag SE is free to download and use with core features including chat, document upload, and encryption. Optional in-app purchases unlock additional capabilities.
SafeRag SE is the personal edition for individual users, available on the Mac App Store with built-in AI models and zero setup. SafeRag is the enterprise edition with HIPAA/GDPR compliance, multi-user RBAC, full audit logging, and is distributed as a direct download for organizations in regulated industries.
No. All AI processing, document analysis, and data storage happen entirely on your Mac. SafeRag SE only uses the network for downloading AI models and App Store purchase verification. Your conversations and documents never leave your device.
On macOS 26 (Tahoe) and later, SafeRag SE can use Apple's on-device Foundation Models. These run through Apple's secure on-device infrastructure — no download needed, instant availability, and completely private.
No. Unlike SafeRag, the SE edition has a built-in AI engine (llama.cpp) that runs models directly. Ollama is entirely optional — you can connect to it if you already have it installed and prefer its model library.
Upload documents (PDF, TXT, MD) and SafeRag SE creates local vector embeddings using sqlite-vec. When you ask questions with RAG mode enabled, the app finds relevant document sections and includes them as context for AI responses. Everything is processed locally.
Yes. SafeRag SE uses AES-GCM encryption for chat messages. Set a passphrase during setup and your conversations are encrypted at rest in the local database. Without the passphrase, message content cannot be read.
SafeRag SE requires macOS 15.0 or later and an Apple Silicon Mac (M1/M2/M3/M4) with at least 8GB RAM. Apple Foundation Models require macOS 26 or later. Intel Macs are not supported.
Yes. As a Mac App Store app, SafeRag SE runs in Apple's App Sandbox. It can only access files you explicitly grant permission to, and it cannot read or modify data from other apps. This provides an additional layer of security beyond the local-only architecture.
SafeRag SE is built with the help of these excellent open source projects.
High-performance LLM inference engine that powers SafeRag SE's built-in AI capabilities.
GitHub MIT LicenseSQLite toolkit for Swift, providing the database layer for chats and documents.
GitHub MIT LicenseVector search extension for SQLite, enabling fast document similarity search for RAG.
GitHub Apache 2.0Native SwiftUI Markdown renderer for beautifully formatted AI responses.
GitHub MIT LicenseSwift framework for creating and extracting ZIP archives, used for document import/export.
GitHub MIT License