--- title: "A design tool I built for myself, now open source" description: "Atelier is a local-first vector design tool I built to test an idea, then set aside for five months. It's on GitHub now, and here is why picking it up again was easy." date: 2026-08-17 image: "/uploads/2026/08/atelier-open-source-design-tool.jpg" categories: - "Design" - "AI" - "Product" --- I've put a design tool I built on GitHub. It's called Atelier, and you can [try it in your browser](https://atelier-glauser.vercel.app) without installing anything or making an account. It's not new. I built it in March to test an idea, then stopped working on it. What's new is that I've started again, and that it's [open source](https://github.com/oskarglauser/atelier) now. ## What it is A focused workspace for the kind of work I actually do. Sketching ideas, drawing vector marks, building moodboards, putting identity systems together. It's local-first, which is the part I care about most. No account, no subscription, no backend. Your projects stay on your machine and nothing is sent anywhere. It runs in a browser tab or as a desktop app on Mac, Windows and Linux. The vector tools are what you'd expect: pen, freehand, path editing, boolean operations, masks and compound paths. Type uses your system fonts or Google Fonts. You can export to SVG, PNG, JPG, EPS and PDF. One thing that isn't standard is CMYK soft proofing, so you can see how your colours are likely to behave in print. Most browser design tools ignore print entirely. If you've ever handed a beautiful screen logo to a printer and got something muddy back, you'll know why I bothered. ## Why I stopped I built it because I had an idea and wanted to see whether it would work. It did. Then I stopped. What came next would have been pricing it, marketing it and talking about it every week for a year. I didn't want to do any of that. You can see the moment in the release history. Version 0.3.0 went out on 22 March, then nothing for nearly five months. It's the pattern I wrote about in [what's worth building when you can build anything](/thoughts/when-you-can-build-anything-whats-worth-building): starting got cheap and finishing didn't. Atelier is one of more than twenty apps I've built with AI and not taken further. Doing that repeatedly is most of the reason I ended up writing [a book about picking the right problem and proving someone wants it before you build](/thoughts/i-wrote-a-book). It comes out on 15 September and you can [pre-order it now](https://www.amazon.com/dp/B0HCR9XMLZ). ## What changed I'm not a developer. I don't write code and I didn't write this. Every line was built with AI, the first version and everything since. I've [worked this way for a while](/thoughts/how-i-vibe-coded-an-ios-app-without-knowing-how-to-code). What I bring is knowing how a design tool should feel. Last week I opened the project again to see what Fable could do with it. Nine releases in eight days, adding masks, compound paths, better text wrapping and desktop builds for all three platforms. It also gained collaboration. Two people on the same network can open the same document and edit it together, seeing each other's cursors live. There's no server in that path, so it keeps working with the internet unplugged. That's the part that surprised me. It's proper systems work, the kind of thing you'd normally hire someone to do, and asking for it was no harder than asking for the pen tool. If you're describing the thing rather than building it, the distance between an easy feature and a hard one has mostly closed. ## A third option For years there were two honest choices for a side project. Commit to it properly, or let it go. Open sourcing was always available, but it felt like taking on a burden rather than putting one down. You're inviting people to use something you've just admitted you won't maintain. That maths has changed. Fixing a bug is a prompt or two away. So there's a third door: publish it, keep it alive at very low effort, and let it be useful to whoever finds it useful. The thing gets to exist without having to become a business. That option barely existed before, and I suspect a lot of good work is sitting on hard drives waiting for someone to notice it's now available. ## What it isn't It's not a Figma replacement and isn't trying to be. There's no SVG or PDF import yet, so you can get work out but not always back in. Collaboration only works between machines on the same network. The Mac builds aren't signed yet, so macOS will say it can't verify the app and you'll have to click Open Anyway in Privacy and Security. And a share link is the permission: anyone who has it can edit that document, and there's no way to revoke it. ## If you use it, I'll keep going It's GPL-3.0, so you can use it, change it and share it, and modified versions stay open too. I'm not promising a roadmap. But if people actually use this I'll keep putting time into it here and there, and that's now a small enough ask that I can mean it. If something is broken or missing, tell me. There are a few other things on my shelf that could go the same way. ## Nothing is really finished Something you built and set down isn't necessarily finished with. It might just be waiting for the cost of caring about it to come down far enough. For a lot of things sitting on a lot of hard drives, it already has.