FileWhisper reads your PDFs, notes and scanned documents and answers questions about them, entirely on your computer. No uploads, no account, and no API key to get started.
Your IndiGo flight 6E798 from Indore to Kolkata departs at 8:30 PM on 2 July 2026 and arrives at 10:40 PM. Check-in closes 45 minutes before departure, so make sure youâre at the counter by 7:45 PM.
The actual app: a clean local web page that opens in your browser. Answers read like a conversation and cite the files they came from.
Free, no API key needed. Private by design ¡ macOS, Windows & Linux ¡ open source.
Watch FileWhisper index a folder of PDFs and scans, then answer questions with citations.
On a Mac, press Command + Space, type âTerminalâ, and hit Enter.
TerminalPaste the install command from above and press Enter. It sets everything up and adds a FileWhisper icon to your Desktop.
Double-click FileWhisper. It opens in your browser. Pick a folder, then just type your question in plain English.
Cloud tools make you upload your documents to someone else's server. FileWhisper keeps everything on your machine.
| FileWhisper | ChatGPT / ChatPDF uploads | NotebookLM | |
|---|---|---|---|
| Where your files go | Stay on your computer | Uploaded to the cloud | Uploaded to Google |
| Account needed | None | Yes | Google account |
| Cost | Free & open source | Free tier, then paid | Free, usage limits |
| Scanned pages & photos (OCR) | Built in | Varies by plan | Limited |
| Cites the exact source file | Every answer | Sometimes | Yes |
| Choice of AI model | Any provider, or the free built-in one | Fixed | Fixed (Gemini) |
FileWhisper works the moment you install it. No signup, no key. Want quicker, sharper answers? Add an API key from any major provider. Itâs completely optional, and itâs saved on your computer and used directly.
A free built-in assistant answers your questions out of the box. Perfect for trying it out and everyday use.
Paste a key from any provider (Groq, OpenAI, Claude, Gemini, or any OpenAI-compatible API) in the LLM Settings panel for noticeably quicker, sharper answers. Totally optional.
Indexing, search and OCR all happen on your machine. Files never leave your computer.
One command. No Docker, no databases, no Python wrangling. Small install, no PyTorch.
PDFs, text and Markdown, plus scanned pages and images via built-in OCR.
Every answer shows exactly which of your files it came from, so you can trust it.
One command on your server. Five minutes later, your visitors get an âAsk our docsâ search box that answers questions from your own pages, manuals and PDFs, and cites exactly where each answer came from.
No vector databases to run, no pipelines to build, no AI team required. FileWhisper indexes your documentation on your own server and serves search from there.
SSH into your web server and paste the install command. That's the whole setup.
curl ⌠| bashA setup page opens in your browser. Point and click through your server's folders to choose the docs, manuals and PDFs to include.
FileWhisper reads your pages and scanned files and builds a search index, right on your server. Your content never leaves it.
Copy the snippet it gives you into your site's HTML. The AI search box appears instantly.
<script>Let developers ask questions instead of hunting through pages. Every answer links back to the exact doc.
Course portals, lab knowledge bases and paper collections become searchable. Students get cited answers.
Policies, handbooks and internal manuals, searchable in plain language on your intranet.
Equipment manuals, standards libraries and government portals, including scanned PDFs via OCR.
FileWhisper Desktop stays free and open source. FileWhisper for Websites is the commercial product for organizations.
For personal sites, small docs and trying it out.
For teams whose documentation is part of the product.
For organizations with compliance and scale needs.
Pricing is indicative while we're in early access. Waitlist members lock in early-adopter terms.
Tell us about your website and where we can reach you when integration is ready.
Special thanks to Prof. Chetan Mistry for suggesting the PC use-case.