100% on-device · No data ever leaves your machine

Chat with your company's documents. Privately.

192.AI runs a full AI assistant on your own computer. Ask questions across your PDFs, Word docs, spreadsheets, and slides — without sending a single byte to a cloud provider.

Available for macOS, Windows, and Linux

Your documents. Your machine. Your answers.

If you need answers from internal files but can't ship those files to OpenAI, Anthropic, or Google — this is for you.

🔒

100% on-device

No telemetry, no document upload, no cloud round-trips. The model, the index, and your files never leave your computer.

📄

Wide format support

PDF, DOCX, XLSX, PPTX, TXT, Markdown, CSV, and images. Scanned PDFs and photos are read via built-in OCR.

👁️

Live folder watching

Drop a file into a watched folder and it's indexed and searchable within seconds — no manual re-index needed.

📌

Citations on every answer

Every response shows exactly which files and pages it came from. The model answers only from your documents.

🧠

Pluggable models

Ships with curated picks — Llama 3, Gemma, Phi, Mistral, Qwen, DeepSeek — or point it at any model Ollama can pull.

🌐

Optional remote access

Securely query your local index from anywhere via an outbound-only tunnel. Documents stay on your machine; only answers travel.

How it works

A complete retrieval pipeline, running entirely on 127.0.0.1.

1

Point it at your folders

A guided setup wizard installs the local AI runtime and lets you pick the folders to index. One click, no configuration files.

2

It indexes locally

Files are parsed, chunked, and embedded into a local vector store on your disk. New and changed files are picked up automatically.

3

Ask anything

Your question is matched against your documents and answered by a local LLM — with source files and page numbers cited.

┌─────────────────┐ ┌────────────────────┐ ┌──────────────────┐ │ Desktop app │ ←→ │ Local backend │ ←→ │ Local LLM │ │ │ │ (your machine) │ │ (your machine) │ └─────────────────┘ └────────┬───────────┘ └──────────────────┘ │ ▼ ┌────────────────────┐ │ Local vector DB │ │ (your disk) │ └────────────────────┘ ✓ Everything on 127.0.0.1 — zero outbound traffic

Frequently asked questions

Does any of my data ever leave my computer?

No. The language model, the embeddings, the vector index, and your documents all live and run on your machine. There is no telemetry and no cloud processing. The only optional outbound connection is the remote-access tunnel, which you must explicitly enable — and even then, documents never leave the machine.

What hardware do I need?

Any modern Mac (Apple Silicon or Intel), Windows PC, or Linux machine with 8 GB of RAM can run the default model. More RAM lets you run larger, more capable models — the app includes a catalogue ranging from fast 3B models to 70B-class models for high-end machines.

Do I need an internet connection?

Only for the initial model download during setup. After that, 192.AI works fully offline.

What file types are supported?

PDF, DOCX/DOC, XLSX/XLS, PPTX/PPT, TXT, Markdown, CSV, and common image formats (PNG, JPG, TIFF, WEBP, BMP). Text in images and scanned PDFs is recovered automatically with OCR.

How is this different from ChatGPT or Copilot?

Those services process your documents on someone else's servers. 192.AI runs the entire pipeline on your own hardware, which means it can be used with confidential contracts, medical records, financial data, and anything else your compliance team won't let near a cloud API.

Is there a free version?

Yes — the free trial is fully functional for evaluation. See pricing for what the Pro license unlocks.

Answers from your files. Zero data exposure.

Download 192.AI and be chatting with your documents in minutes.

Download for free