What it is
Dungeon Mastron is a free, MIT open-source platform for making and playing choose-your-own-adventure games. Four pieces that connect:
- A browser player. Games are single JSON files. You share a game with a link, anyone plays instantly. No accounts, no installs.
- A visual builder. A node graph where you wire branching story pages together without code.
- The AI Companion. This is the strange one. It's not an app, it's a markdown prompt template. You paste it into ChatGPT or Claude, describe the game you want, and the LLM outputs a complete playable game file. Not "AI writes you some text" but an actual structured game with stats, items, wounds, endings that runs in the player. I'm basically distributing software as a prompt.
- A physical console. A 3D-printable Raspberry Pi shell / console where the game cartridges are USB sticks. You print the case, flash the Pi, drop a game file on a USB stick, plug it in. There's something about physically inserting a cartridge you made yourself that a browser tab can't replicate.
The honest state of things
I'm a solo builder in Sweden. This is my passion project, not my business (I run SaaS products for that). Current reality:
- Web player and builder: live and working, being refined
- AI Companion: working (v7.2), gets better with every generated game
- Console: beta. The v1 printed case exists and works on my desk. The Pi engine runs but is not thoroughly tested yet, I'd call its state pending
- Users: roughly zero. Traffic: roughly zero. Revenue: zero by design, there's a Ko-fi tip jar and that's it
I open-sourced the whole thing (MIT) a couple of days ago and put it on GitHub.
Why give it away?
Because the thing this project needs isn't revenue, it's a community. People to play the example games, break the engine, add features, and honestly, redesign the console shell, because anyone can remix the case files. A CYOA platform with one author is a demo. With fifty authors it's a library.
There are also design decisions in here I want feedback on. One example: the engine uses "wounds, not death." Failing a challenge doesn't kill you and end the game. It wounds you, and the story routes around the wound. Failure changes your path instead of erasing your progress. I think it makes branching stories feel much fairer, but I've only tested it on my own games.
One mistake I'll admit upfront
I built this for over a year before doing any marketing at all. Classic maker trap: the building was the fun part, so I just kept building. The site sat live with zero SEO, zero posts, zero sharing. I'm now doing the unglamorous work of writing guides, fixing sitemaps, and posting here, and it's humbling how much of that I could have done alongside the building.
What's next
Getting the console engine properly tested is the big one. Then getting the first strangers, not me, to build and share a game. That'll be the real proof.
The ask
If any of this sounds fun: play a game in the browser, try generating one with the AI Companion, or if you have a 3D printer and a spare Pi, be one of the first to build the console and tell me everything that breaks.
Everything is at dungeonmastron.com and the code is on GitHub (coburn2716/dungeonmastron_app).