Performance Profiling

Aura can profile your game directly in the Unreal Editor, surfacing performance data from PIE sessions and analyzing it for you. It runs performance captures, identifies CPU, GPU, and render-thread bottlenecks, and reports the results directly in chat — no need to leave the editor or open a separate profiler window.

ℹ️ Make sure you're in Agent mode to use Performance Profiling.

⚠️ Performance Profiling is in alpha. Coverage of profiler data is still expanding, so results may vary depending on the specific issue.

Capturing a Profile

Ask Aura to help you profile a session and it will expose the profiler capture controls in chat. From there, you control the start and stop of the capture while you reproduce the issue in PIE.

Examples

• "Help me profile my game — I'm seeing hitches when enemies spawn."
• "My framerate drops when I open the inventory UI. Can you profile it?"
• "I think I'm GPU-bound during combat. Run a profiler capture and tell me what's expensive."

For best results, drive the capture manually:

1. Ask Aura to set up a profiling session.
2. Start your PIE session and get to the moment just before the issue occurs.
3. Press start on the capture controls, reproduce the performance issue, then press stop.
4. Aura will analyze and report what it finds.

💡 The more detail you give Aura up front, the better. A capture can contain a lot of data — telling Aura what kind of issue you're investigating ("frame hitches when enemies spawn," "sustained low framerate during combat," "long load when entering the level") and where to focus helps it direct its analysis.

What Aura Can Identify

Aura can usually pinpoint basic CPU/GPU-bound issues and hitches from a capture, including:

• Whether you're CPU- or GPU-bound on average and during spikes
• Game/Render-thread bottlenecks
• Frame hitches and what's running during them
• Hot functions or systems that show up across the capture window

It does not yet have access to every category of profiler data, so some specific situations — niche subsystems, fine-grained GPU pass breakdowns — may not produce a clear answer. If a capture doesn't tell you what you need, try narrowing the repro and capturing a shorter window focused on the moment in question.

Autonomous Capture

Aura also supports autonomous profiler captures, where it starts and stops the capture itself rather than waiting on you. This mode is intended for use with the verification agent and is highly experimental — it's still finding its footing.

For day-to-day profiling work, we recommend driving the capture manually. You'll get more reliable results because you control exactly when the recording window opens and closes around the issue you care about.

ℹ️ Using Aura through Claude Code or another external IDE? Manual capture relies on the start/stop buttons that live in the Aura chat panel inside Unreal, so it isn't available from external clients. Autonomous capture is the only option in that setup — see IDE/Claude Code.

Workflow Tips

• Be specific about the symptom. "Stutters when an enemy spawns" gives Aura a much better starting point than "the game feels slow." Tell Aura what you saw, when, and what you suspect.
• Capture a tight window. Shorter, focused captures around the exact repro produce cleaner analysis than long sessions full of unrelated activity.
• Tell Aura where to look. If you already suspect a particular system, level, or asset is at fault, mention it. Aura will direct its attention there first.

Current Limitations

We're actively working on improving Performance Profiling. It'll only get better from here!

Profiler data coverage

Aura doesn't yet have access to every type of profiler data Unreal exposes. Basic CPU/GPU/render-thread analysis is solid; more specialized data may not show up in the capture Aura sees.

Autonomous capture is experimental

Autonomous capture is a preview feature and is most useful in conjunction with the verification agent. For best results today, control the start/stop of the capture yourself and reproduce the issue manually.

Best on reproducible issues

Aura works best when you can reliably reproduce the performance issue inside PIE. Intermittent issues are harder to capture and harder to analyze.