@brainit-consulting4Code Dungeon: walk through the workshop where your AI agents are building
Each project has its own chamber. Each workbench has a live terminal. The characters in the dungeon are real coding agents working through your GitHub issues.
Code Dungeon is my local 3D workshop for managing agent-built software: plan the work, watch coding sessions, review tester reports and pull requests, then ship when ready. It works with your own agent subscriptions, and includes a separate pretend-project demo mode. It grew out of Leon van Zyl’s cubefarm and is MIT licensed.
I would value feedback on setup and visibility: can you understand what each agent is doing, and what evidence do you need before trusting a change?
Introduction: youtu.be/RzP2SoLmC6Q
Getting started: youtu.be/Mn74ZSlh78U
User guide: codedungeon-guide.vercel.app/
Source: github.com/brainit-consulting/codedungeon
Free Skool community: skool.com/codedungeon-1002
More videos: youtube.com/@BrainITConsulting
The banner is promotional concept art; the videos show the actual application, using demo projects.
Comments1
Giving each project its own chamber, with a character for every agent working a GitHub issue, makes it easy to see at a glance who is busy and who is stuck. For a founder juggling several agent-built projects, having the tester reports in the same place where you decide to ship is what turns watching the work into trusting it.
1 comment
Giving each project its own chamber, with a character for every agent working a GitHub issue, makes it easy to see at a glance who is busy and who is stuck. For a founder juggling several agent-built projects, having the tester reports in the same place where you decide to ship is what turns watching the work into trusting it.