Qiuner/QCode

What it solves

QCode transforms the traditional, text-heavy AI coding experience—typically confined to chat boxes and terminals—into an explorable, visual world. It aims to make AI-assisted development less intimidating for beginners by turning abstract agent capabilities into "residents" and "places" in a virtual island, allowing users to learn coding through real-world project creation rather than isolated courses.

How it works

QCode uses a Godot-rendered 3D world as its interface, where different AI residents act as entry points for specific capabilities. For example, "Q" serves as the maker companion for project execution, while "Uncle Moss" manages project history. Under the hood, it integrates with the DeepSeek Harness runtime to handle the actual LLM interactions, tool calls, and file system modifications. The system combines a React-based workspace for interaction and a Godot-based world to reflect the AI's working state through animations and speech bubbles.

Who it’s for

It is primarily designed for people new to AI coding who want a guided, low-friction way to start building real projects, as well as developers who prefer a more visual and gamified environment for managing their AI coding sessions.

Highlights

• Gamified Interface: Replaces standard chat interfaces with an explorable archipelago where AI agents are represented as residents.
• Integrated Learning: Combines guided tutorials with real project execution in a shared workspace.
• Resident-Based Capabilities: Assigns specific roles to AI characters (e.g., File Keeper, Project Keeper) to organize agentic workflows.
• Professional Workbench: Maintains an advanced interface for users who need to see full tool calls, conversation histories, and manual approvals.
• DeepSeek Harness Integration: Leverages a pinned DSH runtime for consistent execution and tool management.