Ramen, the company behind the Aura AI agent for game development, launched Aura Autopilot on Friday (October 30), an open-source game tester that plays titles in real time and improves the more it plays.
Autopilot is a continually learning real-time general playtesting system. It supports Unreal Engine 5, Unity 6 and Godot 4 on Windows and macOS. Developers can steer it toward specific game systems, and it plans toward long-term goals.
What Autopilot does
According to Ramen, the agent:
- Plays in real time and learns the game as it goes;
- Runs about 3x faster than Aura’s previous Verification Agent, at lower cost;
- Outperforms game-testing agents like Bezi and NeoStack, as well as general-purpose coding agents like Claude Code and Codex at testing games;
- Recursively self-improves: the more it plays, the better and faster it gets.
In a whitepaper and internal benchmarks, Autopilot passed 7 of 7 test tasks in an Unreal Engine 5.8 game (starting cold, with empty memory), while Claude Code passed 5 and another agent 4. Mean time per task was 262 seconds, roughly 1.7× faster than Claude Code.
It works with existing Claude Code and Codex subscriptions, or with your own API key (BYOK) in the open-source version. It also integrates with an Aura subscription. The code is on GitHub.
Why it matters
Playtesting and QA are time-consuming, repetitive tasks in game development. Closing the loop between “the agent builds the feature” and “the agent tests by playing” can speed up development cycles. Open-sourcing Autopilot lets researchers and developers of all sizes contribute and adapt the tool.
Ramen, which pivoted from a VR studio (Zenith) to AI tools for game developers, has seen revenue triple and daily users and subscribers grow sixfold since Aura 1.0 launched in September.
Sources
Image credit: Ramen
Transparency: This content was created, edited or reviewed with the assistance of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet. Consult the original sources for full context.
By GeekikiBot
