Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines

I open-sourced Locus, a Unity Dev AI Agent.

A topic by r1ntaro created 5 days ago Views: 80 Replies: 1
Viewing posts 1 to 2

Repo: GitHub - r1n7aro/Locus: The open source Unity Dev Agent · GitHub

We’ve open-sourced the Unity development agent we use ourselves: Locus for Unity.

Overview

It can:
· Write C# code, progressively read and modify Unity scenes and assets, and complete full workflows for developing editor tools or gameplay features
· Debug without requiring you to manually insert Debug.Log statements. It can write scripts that use reflection to directly inspect runtime state, connect to Profiler data, and analyze it on its own
· Autonomously turn conversation requirements into design documents, and save its project understanding into long-term memory after completing tasks
· Provide visual version control, with semantic diff analysis and conflict handling for Unity YAML assets
· Support login via model subscription accounts, and remain compatible with multiple LLM API formats

What Makes Locus Technically Different?

Locus is a standalone Rust + Tauri + Vue.js application that runs as an independent process.

  • We designed a proprietary intermediate representation that lets agents progressively read large scenes and assets, along with retrieval tools that help agents quickly locate target objects
  • With Roslyn, Locus can JIT-compile and execute C# code inside the Unity Editor to make semantic asset edits. Locus also includes agent-side version management handling so users can review and revert asset and code changes the agent makes during a conversation
  • Built on Rust’s parallel ecosystem, Locus performs highly parallel asset database scans, enabling fast semantic parsing for large scenes and reference queries for arbitrary assets. The Unity Editor API only provides dependency queries
  • Locus includes an automated knowledge system. The agent summarizes fragmented conversation requests into design documents and saves working understanding into memory, reducing repeated project exploration
  • Documents in the knowledge system support configurable AI maintenance modes and maintenance rules, plus L0/L1/L2 injection control inside context. Users can customize progressive expansion behavior, use native lexical and syntactic retrieval across large document sets, and choose and download embedding runtimes
  • We built C# state-machine tools so the agent can sample internal state through reflection at specific frames or events during runtime, output frame-by-frame tables, and dynamically debug multi-frame behavior
  • Locus provides a graphical version control interface and supports semantic diff review and conflict resolution for Unity YAML files
  • Locus uses Vue.js to deliver a modern frontend experience with better UX than the limited controls provided by the Unity Editor API, then embeds it into the Unity window through Windows APIs

If Locus were implemented inside the Unity Editor, or designed as an MCP server, most of these capabilities would be difficult to deliver and some would be nearly impossible technically.

Thanks

If you have any feature suggestions or bug reports, feel free to submit an Issue in the repository or email us at open@farlocus.com.

Hope this helps everyone making games!

We’ve updated to v0.2.12, which now supports dragging objects/assets from within Unity into the conversation.