Well Extract - AI-powered receipt & invoice extraction for developers

Well Extract — Lightweight AI-powered invoice & receipt data extraction tool for developers. CLI-first, open source, and customizable. Run it locally, choose your AI model (OpenAI, Anthropic, Gemini, Ollama), and get structured JSON from any invoice.

Add a comment

Replies

Best
Hey Product Hunt! 👋 We’ve been quietly building a set of open-source tools under the Well brand — lightweight AI utilities designed to slot straight into developers’ workflows. Well Extract is our first launch. It does one thing well: turn messy invoices and receipts into structured data you can actually use.
 Why we built it Parsing invoices and receipts is boring. Existing APIs often lock you into a vendor, pricing model, or platform. We wanted something open — a tool you can run locally, point at your docs, and get clean structured data out — ready for your pipeline. No subscriptions. No lock-in. No black box.
 What it does 📄 Takes invoice/receipt PDFs or images 🤖 Runs extraction using your choice of AI model (OpenAI, Anthropic, Gemini, Ollama) 🧱 Extracts to your own JSON spec — prompt the structure you need, get output your way 🔗 Pipes straight into your scripts, apps, or data stack ⚙️ Runs via CLI — fast, local, open-source
 Who’s it for 🛠️ Developers building tools around invoices, receipts, accounting 📊 Data teams needing reliable extraction from documents 💻 Solo founders automating boring financial ops 🤓 Anyone who likes lightweight tools they can tweak and control
 Why open source? We believe developers should have the freedom to choose their stack — not be locked into expensive APIs. This is the first “satellite” launch for Well — we’re aiming to ship more useful open AI tools soon.
 We’d love your feedback This is still a young project — we’d love to hear how you’d use it, what features you want, and what we should build next. Star the repo, try it out, or just say hi here. Thanks for checking out Well Extract! 🚀
I’m curious—how does the tool handle different document layouts and formats to ensure consistent extraction accuracy?

 You design the format output model in the prompt you give to the LLM when running OCR. Prompt can be changed depending on the type of document.

Can this tool also create invoices and somehow help with filling out the tax forms? 😅 That is the real pain point (and I reckon that I am not only one).

 Yes! we have another package in the repo that generate receipts or invoices based on a prompt. 👀 check it out:

For the tax forms, not yet, but that's the plan

 Do you have a plan to integrate it into your website (and make it more UI/UX friendly)? 👀 I think that many people who invoice and have different kinds of businesses will not work with a code. 😀

 will do, keep you posted when it's live (and with a PH V2 launch ;) )

 on point!
I've had to scramble through a lot of invoices and forms to file taxes and completely relate with your pain point.

 Thankfully, I do this process once per year, but there are many people/founders who do this on a monthly basis.

 I am trying to solve this real pain point with for reals. Not in the hype space of AI AI fully, but trying to solve a real problem of chaotic personal paperwork management. Eventually, getting to improved taxes as well :)

But damn, great launch man - this is undoubtedly a pain point for Devs too

Congrats on the launch!

 Thanks!

Sharp launch from Well Extract 🧾🤖—dev-friendly extraction made simple.

Well Extract is a game-changer for automating receipt and invoice handling! The AI-powered accuracy saves so much time—perfect tool for devs building finance or expense management apps.