Automate Receipt Logging with Clawshier and OpenAI

What this is, in plain English

Clawshier is a free, open-source “skill” that reads receipts for you. You photograph a receipt, an AI that can read images (OpenAI’s vision model) pulls out the shop name, the category, the total, and the tax, and Clawshier writes those details into expense spreadsheets automatically. The goal is to kill the boring manual typing of expense tracking.

Be honest about the setup: this is not a standalone app you just open. Clawshier runs as an add-on (“skill”) inside OpenClaw, an AI-agent framework you install and run yourself, and it needs a paid OpenAI key to do the reading. The exact install commands live in the project’s repository and change over time. So this page explains what it does and what it is good for, then walks the realistic setup, rather than pretending it is a one-tap tool.

What you can use it for

• Shop bookkeeping. Log daily sales or purchase receipts into a running expense sheet without typing them.
• Freelancer expense tracking. Capture business costs as you go so tax time is not a shoebox of paper.
• Travel and reimbursement claims. Photograph receipts on a trip and have them logged ready for a claim.
• VAT and tax prep. Keep a clean, categorized record of spending and tax paid across the year.
• Any “turn a pile of receipts into a spreadsheet” job you would otherwise do by hand.

Tools you need

• Clawshier (free): the open-source skill that extracts receipt details and updates your spreadsheets.
• OpenAI (paid): provides the AI vision model that reads the receipt image. It is pay-as-you-go (you need credits and a card); light use is cheap, heavy use adds up.
• GitHub (free): where Clawshier (and OpenClaw, the framework it runs in) is hosted.

How it actually works

You set up the framework and skill once, then it is a photo-and-go routine. The realistic path:

1. Install OpenClaw and the Clawshier skill. Get them from GitHub and follow the README to install. OpenClaw is the agent framework; Clawshier plugs into it.
2. Add your OpenAI key. Create a key in your OpenAI account and give it to Clawshier, ideally as an environment variable (a setting on your computer) rather than pasting it into a file.
3. Photograph a receipt clearly. Good light, no glare, the whole receipt in frame. Blurry or shadowed photos are the main cause of bad reads.
4. Run the skill on the image. Point Clawshier at your receipt photo. It sends the image to OpenAI to read.
5. Check the extracted details. It pulls the shop name, category, total, and tax. Dates can be unreliable because receipts mix DD-MM and MM-DD, so glance at those.
6. Let it update your sheets. Clawshier writes the entry into its spreadsheet files (a Summary, an Invoice Archive, and a Monthly sheet), so your records build up automatically.

Words you’ll see, explained

• OpenClaw: the open-source AI-agent framework that Clawshier runs inside.
• Skill: an add-on that gives an AI agent a specific ability, here, reading receipts.
• OpenAI API / API key: the API is the online service that reads the image; the key is your secret pass that lets your tool use it and bills you.
• Vision model: an AI that can look at a picture and answer questions about it.
• Environment variable: a setting stored on your computer, used here to keep your API key out of the code.

Original source

Based on a DEV Community post by fdocr describing Clawshier, an open-source OpenClaw skill that automates pulling expense details off receipts with an AI vision model.

Notes & variations

• Do you even need this? For just a few receipts, you can paste a photo into ChatGPT or Gemini, ask it to read the details, and copy them into Google Sheets yourself. Clawshier pays off when you have a steady stream of receipts to log.
• Free-tier alternative: to avoid OpenAI costs and keep data private, a local vision model (like Ollama with a Llama vision model) can do the reading for free, at the cost of more setup and more computer memory.
• Common pitfall: wrong dates from mixed receipt formats. If you hit this, set the date by hand for those entries.
• Tip for better results: the clearer and flatter the photo, the better the read. For faint thermal receipts, more light and a straight-on angle help a lot.