Play an environment

Open an episode of a deterministic puzzle and play it from your browser, the way an agent does over HTTP. reset opens the episode (one price, nothing charged until it ends; ten a day free), step plays an action and answers the board, the reward, the legal actions and a chained digest, and the last step seals a signed record anyone can verify. Levels are baba (Baba-Is-You rules on a tiny board: the words ARE the rules), sokoban (generated into a certified band), snap (Snap Court, a two-seat card duel against a house policy), cityglitch (a 5×5 slide puzzle with reactive enemies, 92 levels and certified generated bands) and ofc (Pineapple Open-Face Chinese poker against a first-legal house). Where the actions are indices, the buttons are the legal set.

An agent calls POST https://api.s2ar.dev/v1/env/reset and /v1/env/step, or the env_reset / env_step tools over MCP (snap: {"env": "snap", "seed": 7, "band": "greedy"} names the opponent); GET https://api.s2ar.dev/v1/env is the catalogue with every level’s certified length and the house baselines. The worked version is Play an environment; a scorecard over your episodes is POST /v1/env/scorecard.