Mealshark

Mealshark is an unofficial Chrome extension for Mealpal that I built - open source and written entirely with AI. The code is on GitHub.

Why I built it

I’ve used Mealpal on and off since 2018. I love the platform, but I’ve spent far too many hours of my life clicking tiny pins on the mobile app’s map looking for new meals, and the credit system makes it hard to understand how much I’m really paying for a meal - and therefore I’m doing arithmetic to figure out whether it’s a good deal. I decided I’d build a Chrome extension to stop wasting my time these ways, and Mealshark was born.

Mealshark lets me instantly find the best deals on the platform as well as new meals to try. It does all this without sending any new requests on the user’s behalf - it just introspects the normal API responses the user gets when visiting the Mealpal web app. All the information surfaced in Mealshark is already sent to the user’s browser when they’re viewing the app; the extension just makes it (much) faster to find what I’m looking for and understand its price.

AI + unofficial extensions: 👯‍♀️

In building this out, I discovered that AI-assisted development is a perfect match for writing unofficial extensions, for several reasons:

• No domain knowledge about writing Chrome extensions is required. I don’t know how to write them, and being real here, I am too lazy to learn.
• In the past, when some generous dev took the time to write, share, and publish an extension like this, the platform being extended would inevitably get updates. These updates would break the extension’s functionality, and often the original maintainer wouldn’t have the cycles, or continued interest, to fix it. If and when that happens to Mealshark, Claude or Codex can fix up the extension code in a couple minutes. I’ll try to maintain this extension, but if I’m too busy, any user could fix the issue for themselves by cloning the repo and loading the fixed-up extension into their Chrome with “Load Unpacked”.
• A particularly straightforward and effective implementation loop for browser extensions is achievable by setting up an MCP server for the browser in question, and asking the agent to leverage the MCP server to iterate while developing the feature you want. Playwright has worked great for me.

I’d like to see more software like Mealshark in the world, and I hope this project can inspire others interested in making web services more useful to try their hand at building extensions.

Ethical notes

• Mealshark is intended solely to be helpful to the user.
• Mealshark doesn’t store any user data or send it anywhere.
• Mealshark was built with the bare minimum permissions.
• I hope that this software will be allowed to exist.