Test a Self-Playtesting Game AI on Your Scene Before Paying

Before you pay for a self-playtesting game AI, break your own build on purpose. The agent earns a paid tier only when it catches the failures you planted, reports the misses, and stays consistent under parallel load.

Run the checks below on your own scene. They are built for a verification agent that claims to playtest without a developer at the controls.

A self-playtesting tool can look like a faster QA pass; the trap is treating speed as reliability.

Vendor speed is not proof your build is safe

Ramen launched Aura 1.0 on September 9, 2026, as an AI development agent for game developers working in Unity and Unreal Engine.

It launched with Mac support for both engines and availability on the Fab Store.

Aura 1.0's Verification Agent is the sole game-development agent able to run playtests without a human developer controlling the session, and it is now eight times faster.

The agent verifies a test case in under one minute, down from a prior two-to-three-minute duration, and can execute three test cases in parallel rather than one.

Ironmace, the developer of Dark and Darker, a game with more than 20 million players, used Aura to create and validate Blueprints automatically, allowing designers without technical backgrounds to build portions of a new tutorial in half the time.

A faster agent is still a hypothesis until it has been checked against a build you know is broken.

Your scene is the real test

Prepare a small broken build with failures you can name. A missing collision, a broken save, a UI button that does nothing, and a state that should not be reachable are all useful. Keep the list short enough that you can review the agent's output in one sitting.

The benchmark below is the part you keep. Each item is a quick check once the broken build is ready.

• Run a prepared seeded failure. Use a broken build you already made, then run a test case against it. The agent should name the failure or point to the relevant area.
• Check for false-clears. Run the same broken build again and look for a clean pass. A false-clear is the dangerous result, because it lets a bad build move forward.
• Compare caught failures to your seed list. Count how many planted problems the agent flagged. The target is a rate you can live with.
• Compare parallel and serial results. Run a short case set in parallel, then serial, and note any disagreement. Stable output matters more than raw speed when the agent is making decisions about your build.
• Time a verification pass. Measure how long a test case takes on your scene. If the time is too slow for your iteration loop, the paid tier may not save you enough.

If parallel runs disagree, treat that as a stability problem. A verifier that changes its answer under load is harder to trust in a release pipeline.

If the agent misses a seeded failure, do not ask whether it is almost right. Ask whether you would ship the build it cleared. If the answer is no, the agent has not earned the trust a paid tier implies.

Use the vendor's numbers as a floor, not a promise

The price tiers include a $10 monthly Indie tier, a $40 monthly Pro tier, a $200 monthly Ultimate tier, and a $150 upfront Fab Store option that includes a one-year Aura subscription.

The $10 Indie tier discount from beta is now permanent.

Auto mode usage is unlimited, including premium models, with rate limits that most developers will not hit in day-to-day work.

Aura 1.0 runs on Anthropic's Claude, and Ramen's team collaborated with Anthropic's Applied AI engineers during the build, centering on a verification system inside the engine that can test games.

The tiers are easy to compare. The harder question is whether the lower tier gives you enough room to run the same checks repeatedly on your own scene.

If the agent clears a build you know is broken, stop there. If it catches the seeded failures and stays consistent, then the paid tier is worth a trial on a real feature.