What Zengist does
Zengist is a free summarizer that runs in your browser and on your own hardware whenever it can. You drop in documents, pictures, links or text, say what you want, and get one brief. Here is the whole feature list, with the limits stated plainly.
Runs on your device
Desktop Chrome and Edge use the browser's built-in model. Any WebGPU browser can download a private one. Your files stay on your computer.
Free, no account
Nothing to sign up for, nothing to install. Open the page and drop a file.
Any kind of input
PDFs, Word documents, text and Markdown, spreadsheets as CSV, web pages from a link, screenshots and photos of text, or pasted text.
Many sources, one brief
Up to 20 files and 10 links in a run. The brief ends with one line per source, so you can see what each contributed.
Any task, not only summaries
Write what you want: list every deadline, rewrite this plainly, compare these two, turn this into an email. The output keeps the same calm shape.
Reads scans and pictures
Text recognition runs in the browser, so a scanned PDF or a photo of a letter works like any other document.
Nothing is stored
No history, no recent list, no copy on a server. A brief lives in the tab and is gone when you close it.
An honest label on every result
Each brief says where it ran: on your device, or on the capped cloud fallback for browsers that cannot run a model.
What you can drop in
Each kind of input has its own page with the details and limits that apply to it:
- PDFs: Summarize a PDF without uploading it.
- Word documents: Summarize a Word document (DOCX) on your own device.
- Web pages: Paste a link and get a brief of the article.
- Screenshots and images: Drop a screenshot, a photo of a page or a scanned image and get a brief.
- Plain text, Markdown, CSV, HTML, JSON, logs and subtitles are read as text. Paste text straight into the card as well.
What a brief looks like
Every result has the same shape, whichever engine wrote it:
- TL;DR. One to three sentences that answer the task.
- Key points. The substance, in order of importance.
- Dates, amounts and actions. Every one found in the material, so nothing that matters is buried in prose.
- Per source. One line per file or link saying what it contributed.
Short, Medium or Long in settings sets how much of the middle you get. The task you type steers all of it.
Where it runs
Three engines, chosen in this order: Chrome's built-in model on desktop Chrome and Edge, a private model your browser downloads once and keeps if it has WebGPU, and a capped cloud model for everything else. You can pin one in settings. How it works explains each in detail and the comparison page sets it against the alternatives.
Limits
- 20 sources per run, 50 MB per file, 10 links per run.
- Scanned PDFs: 30 pages per file. Normal PDFs: no page limit.
- Cloud fallback: about 240,000 characters per run, roughly 40,000 words, and a daily cap so it can stay free. On your device there is no length cap.
- Not read: password-protected PDFs, .doc files, pages that need a login.
What it is not
It is not a chat, a file store or a research assistant with a memory. It reads what you give it, once, and hands back a brief. That narrowness is what lets it stay free and stay private. Questions are answered in the FAQ; if something is not working, start at help.