Behavior Tree Agent

Aura can create, inspect, and edit Unreal Engine Behavior Tree assets. It can currently handle Selectors, Sequences, Tasks, Decorators, Services, Blackboard keys, and EQS (Environment Query System) queries.

ℹ️ Make sure you're in Agent mode to use the Behavior Tree Agent.

⚠️ The Behavior Tree Agent is in alpha. Save and back up your project regularly.

Creating a Behavior Tree

Switch to Agent mode and describe the AI behavior you want. Aura will create the Behavior Tree and Blackboard assets, wire up the nodes, and save them to your project.

Examples

• "Create a simple Behavior Tree for a patrol enemy with idle and chase states."
• "Create a Behavior Tree for an enemy that wanders randomly, chases the player when in range, and retreats when health is low."

💡 The more specific you are, the better the result. The structure of a Behavior Tree can vary significantly depending on which node types and conditions you want. If you describe only the high-level behavior, Aura may produce something that looks plausible but doesn't work as intended. Describe the specific nodes, conditions, and tree structure you have in mind when you can.

Editing and Inspecting Behavior Trees

Aura can make targeted edits to existing Behavior Trees and explain what a tree does.

Examples

• "Add a decorator to the chase sequence that checks if the player is within 500 units."
• "Add a Blackboard key 'TargetActor' of type Object."
• "What does my EnemyBT behavior tree do?"
• "Add a cooldown decorator to the attack task."

💡 Aura is well-suited for iterative edits. If something isn't quite right, it's usually easier to ask for a targeted fix than to start over.

Workflow Tips

• Start simple, then build up. A basic patrol/chase tree is a good starting point. Add decorators, services, and tasks in follow-up prompts once the structure is in place.
• Name your assets in the prompt. Referencing your Blackboard and Behavior Tree by name (e.g., @EnemyBB, @EnemyBT) helps Aura target the right asset.
• Describe conditions explicitly. Instead of "chase the player when close," say "add a decorator that checks if the player is within 500 units using a Blackboard key comparison."
• Use Blueprint Tasks for custom logic. If you need behavior that isn't a built-in BT node, ask Aura to create a Blueprint Task or Service alongside the tree.

Environment Query System (EQS)

Aura can create and edit EQS query assets directly from the Behavior Tree Agent. Describe the spatial query you need and Aura will build the asset, configure the generators and tests, and wire it into your Behavior Tree.

Examples

• "Create an EQS query that finds cover positions away from the player."
• "Add an EQS query to my patrol behavior that finds valid patrol points in a grid around the enemy."
• "Create an EQS query that finds the nearest enemy to the player."

💡 EQS queries can also be created and referenced from the Blueprint Agent — if you're building AI logic in a Blueprint rather than a Behavior Tree, you can ask Aura there too.

EQS Tips

• Name your generators, tests, and parameter data bindings explicitly. Telling Aura "use a Grid generator with a Distance test that scores based on proximity to the player with a customizable MaxDistance parameter" gives much more reliable results than "find a good position."
• Reference your EQS asset by name in follow-up prompts. If Aura created the query in the same thread, it can update it directly. If the asset already exists, drag it in as context or reference it by name.
• Inspect before you wire. Ask Aura to explain the query structure before connecting it to your Behavior Tree — it's a quick way to confirm the generators and tests are set up the way you intended.

Current Limitations

We're actively working on improving Aura's Behavior Tree capabilities. It'll only get better from here!

Perception System

Aura has limited understanding of Unreal's built-in AI Perception system. If your Behavior Tree logic relies on forwarding perception stimuli (like sight or hearing) to a Blackboard key, you may need to guide Aura or set that part up manually.

Blueprint Tasks and Services

Aura may lean toward implementing behavior in custom Blueprint Tasks rather than using equivalent built-in BT nodes. This isn't wrong, but if you want to use native nodes, be explicit about it.

Agent selection

On some requests, Aura may initially reach for the Python Agent instead of the Behavior Tree Agent. If this happens, clarify in your follow-up: "Use the Behavior Tree tools directly, not Python."