Unreal Engine#

Overview#

Use this skill to connect OpenClaw-style agent workflows to Unreal Engine 5.0+ in a way that survives engine minor-version differences. In the evaluated Unreal Engine source repo, built-in RemoteControl, HttpBlueprint, JsonBlueprintUtilities, PythonScriptPlugin, and WebSockets support are already present, so prefer composition over engine patching.

Prefer these integration paths:

1. Editor automation path: Use Unreal Remote Control for editor-time inspection and automation.
2. Blueprint low-code path: Compose HttpBlueprint + JsonBlueprintUtilities for Blueprint-driven request/response workflows.
3. Project/plugin path: Add a lightweight plugin that exposes OpenClaw-specific transport, auth, session, and task hooks to C++ and Blueprints.

Do not assume a single Unreal API surface works identically across all 5.x releases. Keep compile-time dependencies minimal and isolate version-sensitive code.

Integration strategy#

1. Pick the right control plane first#

Choose based on the task:

• Blueprint inspection, editor object/property changes, calling exposed editor functions, preset-driven tooling
  • Prefer Unreal Remote Control.
  • Best for editor-time orchestration without deeply embedding agent logic into gameplay code.
• Blueprint-first HTTP/JSON workflows without much C++
  • Prefer built-in HttpBlueprint + JsonBlueprintUtilities.
  • Best for quick prototyping and designer-friendly task flows.
• Editor automation where Python is a better fit than C++/Blueprint
  • Prefer PythonScriptPlugin.
  • Best for content tooling, batch editor scripting, and rapid iteration.
• Runtime communication from a packaged game or PIE session, Blueprint utility nodes, project-specific task execution, custom authentication/transport
  • Prefer a project plugin.
• Both editor-time and runtime tasks
  • Use both: Remote Control for editor-side work, built-in Blueprint/Python tooling where appropriate, and a plugin for runtime-safe OpenClaw-specific behavior.

2. Keep version compatibility broad#

For UE 5.0+

• Prefer engine modules with long-lived APIs: Core, CoreUObject, Engine, HTTP, Json, JsonUtilities, WebSockets.
• Wrap version-sensitive code behind helper functions or macros.
• Avoid depending on editor-only modules in runtime code.
• Split runtime/editor concerns into separate modules only if the task actually needs editor extensions.

3. Treat Blueprint support as a first-class requirement#

When the user mentions Blueprints, do not stop at C++ plumbing. Provide:

• UBlueprintFunctionLibrary nodes for one-shot operations
• Optional UGameInstanceSubsystem or UEngineSubsystem for persistent connections/state
• Delegate/event surfaces for async results
• Simple JSON-in/JSON-out fallback nodes when typed schemas are not settled yet

Default architecture#

Use this mental model unless the project requires something else:

• OpenClaw side
  • Orchestrates tasks, optionally talks HTTP/WebSocket, can emit structured JSON task envelopes.
• Unreal Editor
  • Remote Control exposes editor-manageable objects/functions/properties.
• Unreal plugin
  • Provides Blueprint nodes and runtime transport.
  • Receives task envelopes, dispatches to Blueprint/C++ handlers, returns structured results.

Task patterns#

Pattern A: Editor-time Blueprint and asset tasks#

Use when the goal is to:

• inspect or tweak actors/components in the editor
• trigger exposed functions on editor utilities
• update properties on Remote Control presets
• support technical art / virtual production style workflows

Approach:

1. Ensure the project enables Unreal Remote Control.
2. Expose only the objects/functions/properties needed.
3. Have OpenClaw call the Remote Control HTTP/WebSocket API.
4. Keep destructive/editor-mutating actions explicit and scoped.

Good fit:

• level setup helpers
• virtual production controls
• editor utility actors/widgets
• batch property updates

Pattern B: Runtime or gameplay-adjacent tasks#

Use when the goal is to:

• connect a running client/editor session to OpenClaw
• send prompts/tasks/results over HTTP or WebSocket
• make tasks callable from Blueprints
• support packaged builds or PIE behavior

Approach:

1. Add a runtime plugin.
2. Put transport in a subsystem or manager UObject.
3. Expose simple Blueprint nodes.
4. Return structured results and explicit errors.

Good fit:

• AI-driven NPC tooling
• in-editor copilot panels backed by runtime-safe transport
• gameplay testing hooks
• diagnostics and telemetry bridges

Workflow#

Follow this sequence.

Step 1: Identify the integration target#

Determine:

• engine version or version range
• editor-only vs runtime vs both
• whether Blueprints must be supported directly
• transport needs: HTTP, WebSocket, local CLI bridge, or mixed
• whether packaged builds matter

If the source repository is not accessible, say so clearly and proceed with a standards-based scaffold.

Step 2: Inspect the project structure#

For an Unreal project, look for:

• *.uproject
• Source/
• Plugins/
• existing module .Build.cs files
• any existing editor utility widgets, subsystems, or automation scripts

If you are generating a new plugin, prefer placing it under Plugins/<PluginName>/.

Step 3: Choose a plugin shape#

Default to a single runtime module when possible.

Create extra modules only if needed:

• Runtime module for packaged/runtime-safe features
• Editor module only for details panels, menus, custom tabs, or editor-only APIs

Step 4: Expose Blueprint surfaces#

Minimum useful surface:

• Connect/disconnect node
• Send JSON task/request node
• Async result delegate or polling getter
• Availability/status node

If the schema is immature, prefer these stable nodes first:

• ConnectToOpenClaw(Url, ApiKey)
• DisconnectFromOpenClaw()
• SendTaskJson(TaskJson)
• GetConnectionStatus()

Then layer typed nodes later.

Step 5: Keep JSON contracts explicit#

Define small message envelopes like:

json

{
  "id": "task-123",
  "type": "run_blueprint_task",
  "task": "summarize_scene",
  "payload": {
    "target": "/Game/Maps/TestMap"
  }
}

Return results like:

json

{
  "id": "task-123",
  "ok": true,
  "result": {
    "summary": "..."
  },
  "error": null
}

Keep these envelopes version-neutral and avoid leaking Unreal internals into the protocol unless necessary.

Step 6: Handle UE version differences conservatively#

Prefer defensive code:

• Use minimal includes in headers.
• Put module-heavy includes in .cpp files.
• Avoid relying on newer helper APIs when older equivalents exist.
• Gate minor differences with ENGINE_MAJOR_VERSION / ENGINE_MINOR_VERSION checks when required.

Blueprint guidance#

When asked for “all types of tasks including blueprints,” interpret that as at least:

• Blueprint-callable connection nodes
• Blueprint-callable request/response nodes
• Blueprint events/delegates for async completion
• a sample actor/component/subsystem showing usage

Prefer these Unreal patterns:

• UBlueprintFunctionLibrary for stateless convenience methods
• UGameInstanceSubsystem for persistent connection lifecycle
• DECLARE_DYNAMIC_MULTICAST_DELEGATE_* for Blueprint events

Avoid over-engineering the first pass. A minimal plugin that compiles across 5.0+ beats a feature-rich plugin that only works on one minor version.

Remote Control guidance#

Use Remote Control when the task is fundamentally editor-side. Expect HTTP/WebSocket-based control of exposed properties and functions. Do not describe it as a packaged-game runtime automation layer unless the project specifically wires that up.

If the user asks to manipulate Blueprints broadly, separate the request into:

• editing/inspecting Blueprint-owned objects in the editor → Remote Control/editor tooling
• calling Blueprint nodes at runtime → plugin-exposed Blueprint library/subsystem

Suggested repository deliverables#

When scaffolding an Unreal integration, prefer these outputs:

• references/architecture.md — integration decisions and tradeoffs
• references/blueprints.md — Blueprint support patterns
• assets/OpenClawUnrealPlugin/ — drop-in UE plugin scaffold
• optional script(s) to copy/install the plugin into a target UE project

Validation checklist#

Before claiming success, verify:

• plugin descriptor exists and names the module(s)
• .Build.cs includes required dependencies only
• public headers are minimal
• Blueprint node names are clear and async behavior is documented
• no editor-only module is required for runtime-only use
• version claims are phrased as “designed for UE 5.0+” unless actually compiled/tested across versions

Resources#

references/#

Read these when needed:

• references/architecture.md for integration choices and compatibility rules
• references/blueprints.md for Blueprint-facing API patterns
• references/remote-control-notes.md for editor automation guidance
• references/openclaw-api-contract.md for the default /api/unreal/task request/response schema
• references/repo-evaluation-notes.md for the verified target repo structure and integration implications
• references/engine-builtins.md for built-in Unreal systems to compose with before adding custom code
• references/install-strategy.md for engine-source vs project-plugin deployment choices
• references/integration-map.md for mapping OpenClaw tasks onto confirmed engine systems
• references/sample-blueprint-flow.md for built-in Blueprint HTTP/JSON wiring
• references/sample-python-flow.md for editor Python automation patterns
• references/openclaw-server-notes.md for the OpenClaw-side handler boundary
• references/sample-uproject-plugin-config.md for project-level plugin enablement
• references/openclaw-server-handler-example.md for a concrete /api/unreal/task handler sketch
• references/remote-control-extension-points.md for repo-observed Remote Control / WebRemoteControl source touchpoints
• references/webremotecontrol-route-observations.md for concrete route/request/response observations from the inspected engine branch

scripts/#

Use scripts/install_plugin.sh to copy the scaffold plugin into a target Unreal project.

assets/#

Use assets/OpenClawUnrealPlugin/ as the drop-in plugin scaffold for UE 5.0+ projects.