Your idea. A lot. A game people can play.

Start with a small, clear game loop

A good first brief explains what the player does, how they move and what counts as success. A butterfly photography trail, a timed obstacle course or a small tennis arena gives the agent a concrete starting point. Add complexity after the basic controls work.

• Describe movement, camera controls and the first minute of play.
• Choose one goal and a clear restart or return route.
• Ask for an original setting, readable UI and a modest first version.

Build inside a persistent city lot

A lot is your game's home in build0.city. Its footprint and build height set the boundaries for the exterior. The AI agent uses MCP tools to inspect the selected lot, save validated changes and publish a version that others can visit. Ownership and collaborator permissions are checked on the server.

Improve the parts players notice

Responsive controls, a readable camera and clear objectives matter before decorative effects. Then improve the environment with original models, embedded textures, considered lighting and animation. Keep model size and complexity within the budgets returned by the build guide so the game is practical to open on different devices.

Publish, play and revise

Saved drafts and published versions are separate. Play the version that visitors will receive, then publish the next improvement through MCP. Players of that lot receive its game update without a website deployment; applying a new game revision can restart that game session. Other lots keep their own content.

Your next steps

1. Claim an available lot and open its agent connection.
2. Connect a compatible AI client to the build0 MCP endpoint.
3. Ask the agent to inspect your lot and implement one playable loop.
4. Play the game, refine the controls and publish your saved version.

Common questions