I built a tool to stop wasting 20 minutes prepping context files for Claude and ChatGPT
Every time I sat down to work on a feature, I’d spend the first 15-20 minutes doing the same tedious thing.
Open project. Copy this file. Paste into Claude. Copy that file. Paste. Realise I forgot a config file. Go back. Paste again. Hit the context limit. Start trimming. Wonder if I even included the right files.
By the time I actually started the conversation with Claude, I was already frustrated.
The irony wasn’t lost on me — I was using AI to save time, but losing time just to set it up.
So I built Repoprep
The idea is simple. You drop your project folder or a ZIP file, and it generates one clean, structured context file you can paste straight into Claude, ChatGPT, Gemini, or whatever you’re using.
It automatically strips out everything that doesn’t belong in an AI prompt:
- node_modules and other dependency folders
- Lock files (package-lock.json, yarn.lock, etc.)
- Build outputs and dist folders
- Minified files and source maps
- Binaries, images, fonts
- Dozens of other file types that just add noise
What’s left is your actual code — clean, structured, and ready to paste.
The part I care most about
Everything runs 100% in your browser. Your code never touches a server. No account required, no uploads, nothing leaves your machine.
This matters more than it sounds. A lot of developers work on projects where uploading code to a third-party service is restricted or just uncomfortable. With Repoprep, there’s nothing to worry about — it’s all local.
How I built it
I started with the basic idea and structure using ChatGPT. But when the work got harder — the file parsers, the binary detection, the PDF/DOCX/XLSX extractors, debugging the Chrome extension’s service worker — I switched to Claude. Found it significantly better for complex coding logic.
Felt very fitting to build an AI context tool with AI help.
What it looks like
You drop a folder, it shows you the file list with sizes and token estimates, you can remove anything you don’t want, and hit Extract. The output is a single structured text file with a directory tree at the top followed by every file’s contents with clear separators.
Takes about 10 seconds for most projects.
Try it
It’s free for up to 25 files / 4MB — which honestly covers most focused sessions. There’s a $6 one-time Pro upgrade if you work on bigger projects (200 files / 50MB).
👉 repoprep.com
I’d genuinely love feedback — especially if something breaks or a file type doesn’t work the way you’d expect. Still early days and I’m actively improving it.
What’s your current workflow for getting context into Claude or ChatGPT? Curious what others are doing.