Hey @Coding-Solo! 👋
The momentum behind your repo has been incredible to watch; it’s clearly the foundation for anyone looking to bridge Godot with LLMs. I’ve been deep in the trenches building 6NineLives/godot-mcp, and after seeing the recent discussions here, I wanted to share a slightly different technical philosophy I’ve been pursuing.
While many tools focus on the workflow/CLI layer, I’ve been focusing on Deep Engine Agency. My goal was to move past "script editing" and into "live autonomous interaction."
Here’s how I’m approaching the harder engine problems:
- Version-Aware Docs AI: To stop the AI from hallucinating Godot 3 code, I’ve built a system that dynamically queries docs matching the user's exact engine version.
- "Vision" Loop: An auto-screenshot system that feeds the AI a live visual context of the editor or running game state, essential for UI/UX agentic work.
- Self-Healing Debugging: Rather than just reading logs, this bridge captures live stack traces and iteratively proposes/tests fixes within the running project.
-
End-to-End Game Testing: I’ve added simulated inputs and state queries so the AI can actually "play" the game to assert logic and find bugs.
-
UID & Asset Resolution: It natively resolves uid:// paths to absolute files, ensuring the AI never loses track of the project structure.
As someone who has successfully navigated the "open-source scaling" phase, I’d value your perspective. I think there’s a real opportunity here to define how MCP servers should interact with game engines, whether that's through a "lite" CLI approach or a "deep" plugin-based bridge.
Would you be open to a quick look or even just a chat about where you see the Godot/AI space heading? No pressure at all, just wanted to connect with the person who really paved the way for this.
Hey @Coding-Solo! 👋
The momentum behind your repo has been incredible to watch; it’s clearly the foundation for anyone looking to bridge Godot with LLMs. I’ve been deep in the trenches building 6NineLives/godot-mcp, and after seeing the recent discussions here, I wanted to share a slightly different technical philosophy I’ve been pursuing.
While many tools focus on the workflow/CLI layer, I’ve been focusing on Deep Engine Agency. My goal was to move past "script editing" and into "live autonomous interaction."
Here’s how I’m approaching the harder engine problems:
As someone who has successfully navigated the "open-source scaling" phase, I’d value your perspective. I think there’s a real opportunity here to define how MCP servers should interact with game engines, whether that's through a "lite" CLI approach or a "deep" plugin-based bridge.
Would you be open to a quick look or even just a chat about where you see the Godot/AI space heading? No pressure at all, just wanted to connect with the person who really paved the way for this.