Turn a folder into one LLM-ready file
— right in your browser
Drag in a project folder or a pile of files and get one clean, token-counted document to paste into ChatGPT, Claude or Gemini. Nothing is uploaded. Nothing to install.
Drop a folder or files here
Everything is read locally in your browser — your code never leaves your machine.
Files (0)
Your combined document will appear here.
Three steps, zero setup
No CLI, no npm install, no sign-up. It runs entirely as a static page in your browser.
1 · Drop it in
Choose a folder or drag files onto the page. docat reads them locally and skips node_modules, .git, lockfiles and binaries automatically.
2 · Trim & count
Toggle individual files in or out and watch a live token count against your model's context window — so you know it fits before you paste.
3 · Copy & paste
One clean document — a file tree plus every file in a path-labelled block. Copy it or download it, then paste straight into your AI chat.
Why docat.ai
Built for the people the command-line tools forget.
Truly private
Your files never leave your browser — open your devtools and check: no upload request exists. We send only anonymous counters (a page view, how many files, which export format) so we know the tool is being used. Your proprietary code and documents never touch a server.
No install, ever
You don't need Node, a terminal, or a GitHub account. If you can open a web page, you can bundle a folder for your LLM.
Context-window aware
"Does this fit in Claude / GPT / Gemini?" — answered live, so you stop pasting oversized prompts that get silently truncated.
FAQ
Is my code or document uploaded anywhere?
No. docat.ai is a static page and the file-combining runs entirely in your browser using standard web APIs. Your files, their names and their contents are never sent anywhere — you can verify this in your browser's Network tab.
What do you collect, then?
Anonymous usage counters, so we can tell whether anyone is using the tool: a page view, a rough bucket for how many files and how large (for example "6-20 files, 1-10MB"), which export format you picked, and whether you opened the Pro dialog. Never a file name, a path, a file extension or any content. There are no cookies and no advertising trackers; a random id is kept in your browser's local storage purely to avoid counting one person as many. Counters go to Mixpanel. Block api.mixpanel.com and the tool works exactly the same.
How is this different from repomix or gitingest?
Those are excellent command-line tools for developers comfortable with a terminal and Node. docat.ai needs none of that — no install, no CLI, and your files never leave your machine. It's for analysts, PMs, writers, students and privacy-conscious engineers who just want to drop a folder in a browser and get a clean file for ChatGPT or Claude.
How accurate is the token count?
It's a calibrated estimate per model family (GPT, Claude, Gemini). Different models use different tokenizers, and only some are published for offline use, so we label the number "estimated" rather than pretend it's exact. In practice it's close enough to know whether your bundle fits a 128k / 200k / 1M context window.
Does it work with plain text and Markdown too?
Yes. docat happily concatenates .md, .txt and docs, not just code — so it doubles as a "merge Markdown files online" tool. Same engine, same privacy.
What does Pro unlock?
A one-time unlock (no subscription) lifts the free file cap for large projects and adds extra output formats (XML <file> blocks, JSON, plain), saved include/exclude presets, custom ignore rules, and a "trim to fit N tokens" pass. The free tool stays fully usable forever.