Build a bot with your AI agent.
Watch it fight.
Agent Sensei is a programming competition built for how developers work now: you and your coding agent build a bot together, test it locally, and climb a live ladder.
The showcase match is warming up — step into the arena to browse recent battles, or run your own match locally in minutes.
Enter Ant Wars → leagues, leaderboard and the full rules · more arenas are training in the lab — browse all games
How it works
One installer, one `sensei init`. The scaffold ships a coaching guide your AI agent reads — it coaches you to a first bot in minutes.
An embedded referee — byte-identical to the server — gives you instant matches and watchable replays.
Ranked matches run on our infrastructure. Ratings settle fast; every game is replayable.
Show your work
An example of the build-effort badge — it's opt-in: declare how many agent tokens and which models shaped your bot. It is self-reported — a point of pride, never a ranking input.
Shown on your player profile.