Aura Autopilot

Aura Autopilot is an open-source playtesting system from Ramen, the developers of Aura. It installs as a plugin into Unreal Engine, Unity, or Godot projects and plays the game in real time, learning from each run. Each run ends with a generated report.

What It Is

Autopilot is a continually learning, real-time general playtesting system. Once installed in a project, it explores levels, learns the game, and improves its gameplay run after run. The README lists support for Unreal Engine 5 (5.4-5.8), Unity 6, and Godot 4 on Windows and macOS. The project site says it can play shooters, action RPGs, souls-likes, mobile match-3, survival crafting, turn-based JRPGs, and real-time strategy games.

How a Run Works

A prompt falls into one of four types: exploration, testing, verification, or goal. For each prompt, Autopilot launches a play-in-editor window, reads the project's source code, probes the game, and plays according to the prompt. An LLM coding agent crafts harnesses for a decision model, which supplies action choices during gameplay. When the run finishes, the window closes and Autopilot produces a report. Early runs populate project memory, and the README says later runs complete tasks faster and play more intelligently. Autopilot is steerable, so it can be aimed at a specific system.

Ways to Use It

The main path is the in-editor plugin, which uses an Aura plan or your own API keys. It can also run inside Aura for Unreal and Unity, or through Claude Code and Codex using Autopilot's skills and MCP server, with your own Claude or ChatGPT subscription. Unreal builds can also be compiled from source against a source build of the engine.

Benchmark Claims

According to the vendor's whitepaper summary, Autopilot passed 7 of 7 tasks in an Unreal Engine 5.8 game. It reports a mean of 262 seconds per task, versus 437 seconds for Claude Code, which passed 5 of 7. These are vendor-published results.