How to use Claude to build a Unity bullet hell
Claude Code built a Unity bullet hell using scene and play-mode tools. Human code and asset review supported an itch.io release; Steam remained future work.
loopedin.games
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Build a Unity two-button bullet-hell prototype within the specified run length and palette. Inspect scene hierarchy through MCP, implement power-up and boss states and check actual play-mode output. Keep generated sprite and sound choices for my review before a playable release.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- LoopedIn Games
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Version 0.1.0 was playable on itch.io after code and asset review. A Steam release was still planned; coverage and speed claims were not independently verified.
- Human involvement
- The developer reviewed code, selected and postprocessed sprites and layered sound effects while judging the game loop.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- The developer fixed the two-button loop, run length and palette before implementation.
- Claude used Unity MCP to inspect scenes, implement systems and check play-mode behavior.
- Human code and asset review preceded the playable itch.io version 0.1.0.
LoopedIn Games used Claude Code with Opus to build Void Balls, a two-button roguelite bullet hell in Unity. The developer constrained the design to a short run and a small color palette rather than leaving the agent to invent an unlimited project.
Unity MCP let Claude inspect hierarchy and components, modify scene objects and enter play mode. The agent implemented C# systems for card power-ups, boss states and the runtime loop. Specialized roles covered design and QA, but the development account also describes human review of the code rather than accepting every generated change automatically.
Visual and audio assets used separate tools. The developer reviewed generated sprites for consistency and postprocessed their edges and transparency. Sound effects needed selection and layering to fit the action. Those steps remained distinct from the agent's ability to create a working Unity scene.
The concrete outcome was a playable version 0.1.0 on itch.io, with a later Steam release still planned. The source's value is its actual engine-feedback loop and constrained game scope, not a claim that the entire creative process became automatic. Runtime checks, code review and asset decisions all contributed to the release, and the article does not establish an independently audited coverage or speed comparison.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Developer blog · LoopedIn Games
Frequently asked questions
What did Unity MCP expose?
Scene hierarchy, components, object edits and play-mode feedback.
What remained human asset work?
Sprite consistency, transparency cleanup and sound-effect selection and layering.
Was the Steam version released?
The article reports itch.io version 0.1.0; Steam remained a later plan.
More details
About this account
The developer fixed the two-button loop, run length and palette before implementation. Claude used Unity MCP to inspect scenes, implement systems and check play-mode behavior.
Original developer account opened. v0.1.0 was playable on itch.io; Steam release remained future work. Near-total test coverage and development speed are not independently verified. Suggested prompt written by Agents In Use; no original task prompt was published in the verified account.
Credit: LoopedIn Games — Developer blog
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Unity 6 · Unity MCP · Replicate · ElevenLabs
Bullet hell · Playable release
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.