For years, my music production lived in Jira. Every song a ticket, every mixdown a status change, every “does this still sound good in the car?” a comment nobody would ever find again. It worked. Sort of. The way a Swiss Army knife works as a screwdriver.
Three evenings with Claude Code later, I don’t need it anymore. Meet Foton Kanban: a native macOS kanban app built for exactly one use case — mine.
What it is
One card is one song. Four columns: open, in progress, in review, done. Whatever sits at the top of a column is what I work on next — position is priority.
Each track carries its production phase as a badge: Jam Session, Arrangement, Mixdown, FX finalizing, Mastering. And there’s exactly one automatic rule: when I drag a track from review to done, the review counts as passed. Before mastering, the track moves up one phase and goes back to work. After mastering, it’s finished.
The code is open source under GPL-3.0: github.com/andremotz/foton-kanban
Why not just use Jira?
Because Jira doesn’t know what a studio is. These are the things I actually wanted — and no big vendor will ever build them for me:
- A listening checklist per track. Car, AirPods, headphones, studio, and the master heard 45° off-axis from the speaker. The card shows
2/5while a track is in mastering. If you produce music, you know why this matters. If you sell project management software, you’ll never care. - Phases as an attribute, not a workflow. A track that fails review goes back to “in progress” and keeps its phase. No custom workflow scheme, no admin console, no consultant.
- My data in plain Markdown. One file per track, in a folder I choose — Nextcloud, iCloud, or a Git repo. No database, no server, no account. The data will outlive the app, and that’s the point.
That last one is the real anti-SaaS argument. Every board action touches exactly one file. There’s no central index for two machines to fight over. And if I stop using Foton Kanban tomorrow, I still have a folder full of readable text files.
I even wrote a Jira importer, so moving out took one command.
How it was built: three evenings, three iterations
I built Foton Kanban with Claude Code. Not as a one-shot prompt, but the way I’d run any software project — just with a much shorter feedback loop.
- Spec first. Before any code, I wrote down the data model and the reasoning behind it. That document became
SPEC.mdin the repo. It’s the most important file in the project, because it tells the AI why, not just what. - Then iterate. Three evenings, three rounds: build, use it on real tracks, notice what’s missing, refine.
The skill that mattered wasn’t writing Swift. It was knowing what I wanted, and being precise about it. Architecture decisions like “one action, one file” came from me. The typing came from Claude.
Is SaaS dead?
Not all of it. But the long tail is.
For two decades, the deal was simple: building your own tool was expensive, so you rented a generic one and bent your workflow to fit it. You paid for a thousand features to use twelve. You accepted the vendor’s data model, their pricing changes and their roadmap.
AI-assisted development breaks that math. When a tailored tool costs three evenings instead of three months, the question flips. It’s no longer “which product fits me best?” but “what exactly do I need?” — and that question has always had a better answer than any product catalog.
I think we’ll see a lot more of this: small, personal, purpose-built software. Not products. Tools. The kind you build for yourself, the way a carpenter builds a jig for one specific cut.
When you should still pay for SaaS
I work in enterprise IT during the day, so let me be honest about where this stops working:
- Teams. Foton Kanban is single-user by design. Permissions, audit trails and twenty people editing the same board are real problems that vendors solve well.
- Compliance and support. If something breaks at 2 a.m., I’m the support hotline. For a hobby studio that’s fine. For a regulated business it isn’t.
- Maintenance. Custom software is a pet, not a purchase. Someone has to keep it alive when macOS or the toolchain changes.
So the rule of thumb: if the tool is core to your workflow and only you use it, build it. If it’s shared infrastructure, buy it.
Take it, fork it, make it yours
Foton Kanban isn’t meant to be your next productivity app. It’s meant to be a template — proof that your own tool is closer than you think. Fork it, rip out the studio stuff, and build the board your work needs. It’s the same itch that made me build FCATS instead of buying an off-the-shelf camera app.
If you want to run it as is: you’ll need macOS 14+ and Xcode to build it. The README has everything else, including a sample board to play with.
And if you build your own version of something you used to rent — let me know in the comments. I’d love to see it.