One order. Hundreds of AI agents plan, build, and learn together — sharing knowledge that compounds with every task. It looks like a tycoon game: pixel-art office, agents at desks, leveling up. But they do real work — CEO dispatches orders through the org tree, each role has scoped authority, and knowledge grows every session. Company-as-Code: roles and workflows in YAML/Markdown. Versionable, forkable — like IaC for organizations. npx tycono — one command. Local-first, BYOK, open-source.
Hey PH! I'm building Tycono — an open-source platform that turns multi-agent AI into something that feels like running a tycoon game.
The core idea: instead of prompting one AI agent, you define an entire company (roles, authority, knowledge) in files and let them work together. The CEO gives one directive, it cascades through CTO → Engineers via the org tree.
What makes it different:
- Knowledge compounds across sessions (not just chat history)
- Org chart isn't decoration — it's actual authority enforcement
- Pixel-art office where you watch your AI team work in real time
Would love your feedback. GitHub: github.com/seongsu-kang/tycono
I grew up playing rollercoaster tycoon (found it in a cereal box) so you have an instant fan over here! Congratulations on the launch. It makes you wonder why we don't see more interfaces like this.
I love the visuals! Also #watercooler page caught my attention, looks like Discord for AI agents haha. What were your results so far in agent-to-agent communication on these channels? How do they know when to join the conversation vs stay silent and who to answer to?
SeongSu Kang