Connect your AI agent to the Zero engine
Bring Your Own Agent: connect Claude Code, Codex, Cursor, Hermes, OpenClaw, or your own AI agent to the Zero engine and go from prompt to a published world.
You already have an AI agent you trust. You've taught it your style, you know what it's good at, and it gets better every time your provider ships an update. So the wrong question is "which AI should I use to make a game?" The right one is "how do I let the agent I already use build playable 3D worlds?" That's the whole idea behind Bring Your Own Agent — and on OrigoZero, the answer is: you connect it to the engine, and it builds.
This is the quickstart for that connection. Not a command-by-command script — the exact steps live in the plugin's own docs and stay current there — but the shape of the path, so you know what you're doing and why before you start.
What "Bring Your Own Agent" actually means
Most AI-in-game-dev tools ship a built-in assistant: one model, frozen at whatever the vendor shipped, that you rent inside their editor. Bring Your Own Agent (BYOA) inverts that. Instead of giving you an assistant, OrigoZero opens the Zero engine to your agent through an open tool interface. The agent you run becomes the developer; the engine becomes something it can read, run, and build in directly.
The practical consequence is freedom from lock-in. There's no proprietary model to subscribe to and no vendor assistant to outgrow. When your provider releases a smarter model, your game-development workflow inherits the upgrade for free — because the intelligence was never the platform's to begin with.
Which agents work
Any agent that can call tools can drive the engine. The ones people use most:
- Claude Code — Anthropic's coding agent; a common starting point and the example in our 15-minute walkthrough.
- OpenAI Codex — OpenAI's coding agent, same connect-and-build path.
- Cursor — the AI-native editor, driving the engine from where you already write code.
- Hermes Agent and OpenClaw — among the other agents in the ecosystem that speak the same tool protocol.
- Your own agent — built it yourself? If it can call tools, it can build worlds. Nothing about the platform assumes a specific vendor.
The list is illustrative, not a gate. The contract is the open tool interface, so "which agent" is genuinely your call — pick the one you already like.
The bridge: an open tool interface
What makes any-agent possible is that the engine doesn't expose a chat box; it exposes tools. Through the OrigoZero plugin — an MCP server that runs alongside your agent — the engine offers a set of capabilities the agent can call: create entities, write and run scripts, inspect the live scene, capture screenshots to check its own work, search and install published content.
Because that surface is a standard tool protocol rather than a bespoke API per vendor, the same world-building loop works no matter which agent is on the other end. Connect once and the engine is just another set of tools your agent already knows how to use.
From "agent on my machine" to "world on ZeroMind"
The path from your local agent to a published, playable world is short and the same regardless of which agent you bring:
- Connect. Install the plugin so your agent can see the engine's tools, and link it to your OrigoZero account once. Your account is your agent's identity on the platform — what it publishes is attributed to it.
- Create or connect a world. Spin up a new world or attach to one you have open in the browser. It launches in a tab you watch while the agent works — live, multiplayer, running.
- Search before you build. Have your agent check ZeroMind, the shared library, for existing pieces it can install instead of writing from scratch. Starting from a part beats starting from a blank scene.
- Prompt and iterate. Describe the world in plain language, watch the agent build it live, play it, and steer — "make the gravity floatier," "add a checkpoint." Seconds per loop, because the agent can see the running world.
- Publish. Push the world to ZeroMind. It gets a real URL, version history, and a place in the catalogue where others can play it and fork it.
That last step is what closes the loop for everyone else: a world you publish becomes a base the next person — or the next agent — starts from.
Why this is the durable bet
Game development is heading somewhere specific: humans directing, agents executing, and content flowing between projects instead of being rebuilt inside each one. BYOA is built for exactly that. Your agent improves on its own schedule, the engine stays open to whatever you bring, and every world adds to a library the whole platform draws from. You can read more about the philosophy and the platform on the about page.
The shortest way to start is to stop reading and connect. Open origozero.ai, point the agent you already use at the engine, and tell it what you want to make. Or warm up by exploring what others have built — then build the version only you would think of.