A sketch-to-prompt tool for Material Design just hit 6,800 GitHub stars — but who's behind it?
M3E Canvas lets you doodle an app screen and turn it into an AI coding prompt, and it's racking up stars fast — though rapid GitHub growth isn't proof of much on its own.
A small open-source project called M3E Canvas has picked up 6,825 stars and nearly 700 forks on GitHub in a fortnight — a growth rate that would normally take a mature developer tool months to reach. The pitch is simple: sketch a Material 3 Expressive app screen in your browser, link the screens together, tap through a rough prototype, then export the whole thing as a prompt for an AI coding assistant like Claude Code, Codex, Gemini CLI or Cursor.
What the tool actually does
According to the project’s GitHub page, M3E Canvas is a TypeScript web app that lets someone rough out a UI for Google’s Material 3 Expressive design language, wire the screens into a navigable flow, and then generate a text prompt describing that design. That prompt is meant to be handed to an AI coding tool, which then writes the actual app code. It’s essentially a bridge between “I drew this on a whiteboard” and “an AI wrote the front end”, aimed at people who like to design visually before typing a single line of code. The repository is MIT-licensed, includes documentation translated into Japanese, Chinese and Korean, and — notably — ships with AGENTS.md and CLAUDE.md files, configuration aimed squarely at AI coding agents rather than human contributors, which fits the project’s stated purpose.
So who is actually behind it, and does it work?
This is where a healthy dose of scepticism belongs. The account behind the project, lnkiai, isn’t a recognisable name in Google’s Material Design ecosystem or a known open-source foundation — it presents as an independent maintainer. GitHub star counts are also a notoriously unreliable measure of real-world usage: they can be driven by genuine grassroots interest, by a project getting featured on a popular newsletter or Hacker News thread, or in worse cases by coordinated or automated inflation. Nothing in the repository itself proves which of these applies here, and there’s no independent benchmark, review, or third-party writeup confirming that the sketch-to-prompt pipeline reliably produces usable code rather than a rough starting point that still needs heavy editing.
It’s also worth being clear about what the tool doesn’t do. It doesn’t write finished apps — it generates a prompt that another AI tool then has to interpret and turn into code, which introduces the same unpredictability as any AI coding assistant: results can vary hugely by tool, model version and the complexity of what you’ve sketched.
Who this is actually for
This is a niche developer and designer tool, not something that affects the average phone or PC user. If you’re not building software with an AI coding assistant, nothing here changes your day. For developers and hobbyists already using tools like Claude Code or Cursor, it’s a potentially handy time-saver for turning a visual idea into a starting prompt — but it’s still an early-stage, unaudited project with a fast-growing but unverified fanbase, not an established or officially endorsed Material Design tool.
The takeaway
M3E Canvas is a genuinely interesting idea — sketch first, prompt second, let the AI fill in the code — and its star count shows plenty of developers are curious. But curiosity isn’t validation. Treat the growth as a signal worth watching rather than proof the tool is essential, try it on a throwaway project before trusting it with real work, and don’t assume popularity on GitHub means quality, safety or longevity.