Blueprint Analyst Documentation
Blueprint Analyst Documentation
Blueprint Analyst is an AI-powered assistant that lives directly inside the Unreal Engine editor. It helps you understand, debug, and get insights on any visual script in your project — without breaking your workflow.
Instead of taking screenshots and asking a chatbot in a browser, you get instant in-editor answers about your Blueprints, Materials, Behavior Trees, and Widget Blueprints.
The most comprehensive AI copilot ever built for Unreal Engine.
Generate Blueprints, animations, materials, Niagara VFX, sequencer cinematics, UMG widgets, PCG worlds, behavior trees, and more from natural language. Analyze your project, talk to your engine with voice, and run concurrent AI agents inside the editor.
💡 Try before you buy with our free 5-day trial, and watch the Overview video.
⭐ Featured on FAB (Feb 5th–9th, 2026) AND (Apr 13th–15th) ⭐
The very first AI Copilot to emerge for Unreal Engine 5 back in 2025, and still the most complete toolset on the market today. This plugin pioneered AI-assisted game development inside the editor and continues to push the boundaries of what's possible. Join the fastest-growing community in the UE ecosystem, trusted by thousands of developers worldwide.
- Brand-new HTML/JS chat UI (replaces the old Slate window).
- 5 reusable API key slots — keep multiple providers / keys configured side by side.
- 5 providers supported: OpenAI, Anthropic (Claude), Google Gemini, DeepSeek, and one Custom (OpenAI-compatible) slot that covers Ollama / LM Studio / OpenRouter / llama.cpp / vLLM and any other OpenAI-compatible endpoint.
- Bring-your-own-key: no shared trial key, no per-token usage cap.
- Vision support — paste a screenshot directly into the chat with Ctrl+V (works with vision-capable models from OpenAI / Anthropic / Gemini).
- Modern context extraction: deeper Blueprint summaries (variables, components, function graphs, perf issues, complexity score), recursive node walking through collapsed graphs and macros, and Behavior Tree / Material support.
- Install & enable the plugin from FAB (or drop the folder into
YourProject/Plugins/). Restart the editor. - Open the UI: Window > Blueprint Analyst (or click the toolbar icon).
- The first time you open it, a welcome screen explains setup. Click Open Settings.
- In the Settings overlay, pick Slot 1, choose your provider, paste your key, click Save, then click Set Active, then ← Back to Chat.
- In the Content Browser, select any Blueprint (e.g.
BP_ThirdPersonCharacter). - Back in the Blueprint Analyst window, click + Full Asset Context. A chip appears confirming the context is attached.
- Type “Summarize this Blueprint’s purpose.” into the input box, hit Ctrl+Enter (or click Send ▶).
- The AI returns a structured Markdown breakdown of the asset.
That’s it.
Context attachment buttons
- + Full Asset Context — analyses the asset currently selected in the Content Browser. For Blueprints this includes parent class, interfaces, widget hierarchy (UMG), components, variables, function-graph walks, a Tick-reachability performance scan, and a complexity score. Materials get expression-graph extraction with input/output wiring. Behavior Trees get tree-structure extraction with decorators and services listed inline.
- + Selected Nodes Context — analyses only the nodes currently selected inside an open editor. Works in: Blueprint Editor, Animation Blueprint Editor, Widget Blueprint Editor, and Behavior Tree Editor. Use this for precision debugging — “Why is this not working?” with just the relevant graph excerpt attached.
- + File — pick a
.txt,.md,.cpp,.h,.cs,.json, or.logfile and attach its contents as context for the next question. Great for sharing crash logs, code snippets, or external docs. - Ctrl+V (paste image) — paste a screenshot directly into the chat input. The image goes in the next message. Requires a vision-capable model (OpenAI, Anthropic, Gemini work out of the box).
Chat management
- + New Chat — start a fresh conversation.
- Chat list (left) — every conversation is saved per-project under
Saved/CodeExplainer/. Click a chat to switch to it. Hover for inline actions:- ✎ rename
- ⇩ export as Markdown
- ✕ delete
- 🗑 Delete all chats — wipes the entire chat list (with confirmation).
- ⇩ Export — saves the current conversation to
Saved/Exported Conversations/<title>.md(usetxtfor plain text — see export-format option in the dropdown).
Send / Stop
- Send ▶ (or Ctrl+Enter) — sends your message.
- ■ Stop — shows up while the AI is generating; click it to discard the in-flight response.
All five providers work the same way in the Settings overlay:
- Click the ⚙ Settings button in the header (or the Open Settings button on the welcome screen).
- Pick one of Slot 1–5.
- Choose a Provider from the dropdown.
- Paste your API key.
- (Optional) Adjust the model name for that provider.
- Click Save, then Set Active if it isn’t already.
- Click ← Back to Chat.
Your keys are stored in plaintext JSON under %LOCALAPPDATA%\BlueprintAnalyst\api_keys.json (cross-project — set once, use everywhere). Treat this file like a secret.
OpenAI
- Get a key: platform.openai.com/api-keys
- Default model:
gpt-4o-mini(cheap, fast, vision-capable). - Alternatives:
gpt-4o,gpt-4.1,gpt-4.1-mini,o3,o4-mini, etc. - Billing must be enabled on your OpenAI account.
Anthropic (Claude)
- Get a key: console.anthropic.com
- Default model:
claude-sonnet-4-5(replace with the latest Sonnet/Opus/Haiku name your account has access to). - All Claude models support vision.
Google Gemini
- Get a free key: aistudio.google.com/apikey
- Default model:
gemini-2.5-flash(free tier covers casual usage). - Alternatives:
gemini-2.5-pro,gemini-2.5-flash-lite.
DeepSeek
- Get a key: platform.deepseek.com/api_keys
- Default model:
deepseek-chat. - Alternatives:
deepseek-coder, plus any newer model names listed in the DeepSeek docs. - Text-only — DeepSeek does not currently support image input. Don’t paste screenshots when DeepSeek is your active slot.
Custom (OpenAI-compatible) — local LLMs & alt providers
The Custom slot speaks the OpenAI Chat Completions API and accepts any endpoint that does the same. /chat/completions is auto-appended if you don’t include it.
| Tool | Base URL | Notes |
|---|---|---|
| Ollama | http://localhost:11434/v1 | Run ollama serve. API key field can be left empty. Model name = the local model tag, e.g. llama3.1, qwen2.5-coder:7b. |
| LM Studio | http://localhost:1234/v1 | Enable the local server in LM Studio’s “Local Server” tab. Model name = the identifier shown in LM Studio. |
| llama.cpp server | http://localhost:8080/v1 | Start with ./server -m ./models/<file>.gguf --host 0.0.0.0. Model field accepts any string. |
| vLLM | http://<host>:8000/v1 | Match the model name to whatever you passed via --served-model-name. |
| OpenRouter | https://openrouter.ai/api/v1 | API key required. Model field = the OpenRouter model id, e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet. |
| Together / Groq / DeepInfra / Anyscale | Their /v1 base URL | Paste their API key. Model = their model id. |
If you have a self-hosted endpoint that requires an Authorization header, paste the bearer token into the API key field. If it doesn’t require auth, leave the API key empty.
Open Settings → Themes to switch between built-in presets (Dark default, Midnight Blue, Forest Green, Rose) or pick a custom accent colour. The new accent applies on the next chat reload.
I'm getting an error when I send a message.
9 times out of 10 it's the API key. Re-open Settings, double-check the key for the active slot, and make sure billing is enabled with that provider. If you're on a Custom slot pointing at a local LLM, make sure the local server is actually running and reachable from the editor.
Does it work offline?
With cloud providers, no — those need internet. With the Custom slot pointed at a local LLM (Ollama / LM Studio / llama.cpp / vLLM), yes, fully offline.
Which model is "best" for explaining Blueprints?
For depth-of-reasoning: claude-sonnet-4-5 or gpt-4o. For speed and cost: gpt-4o-mini or gemini-2.5-flash. For free + privacy: a local model via Ollama (qwen2.5-coder:7b or llama3.1 are good starting points).
Where are my exported conversations saved?
[YourProject]/Saved/Exported Conversations/. A toast notification with the full path appears after each export.
Where are my chats stored?
[YourProject]/Saved/CodeExplainer/conversations.json (manifest) plus one JSON file per chat. Per-project — chats don't follow you to a different project.
Where are my API keys stored?
%LOCALAPPDATA%\BlueprintAnalyst\api_keys.json on Windows, ~/Library/Application Support/BlueprintAnalyst/api_keys.json on Mac, ~/.config/BlueprintAnalyst/api_keys.json on Linux. Cross-project — set once, use in every project.
I'm upgrading from the old 1.x version. Will I lose my saved key?
No. The first time 2.0 launches it migrates your old key from the BlueprintCopilot ini section into Slot 1 automatically.
The AI's answer is wrong / shallow.
AI isn't infallible — rephrasing the question or attaching more context (more nodes, the whole asset, plus a code file) usually helps. Switching to a stronger model (Claude Sonnet, GPT-4o, Gemini 2.5 Pro) helps even more.
Can I add a custom system prompt?
Not in the UI yet. Coming in a future release.
How big a conversation can it handle?
The plugin caps each request at ~320,000 chars of history (roughly 80k tokens). Older turns are silently trimmed. Provider context windows still apply on top of that.
Can I use the same key across multiple projects?
Yes — keys are stored per-user, not per-project.
- Discord: discord.gg/w65xMuxEhb
- More plugins from BlueprintsLab: fab.com/sellers/BlueprintsLab
- Bug reports / feature requests: post on Discord or use the FAB review page.
- FAB Marketplace Listing: Blueprint Analyst on FAB
v2.0
- Rebuilt UI in HTML/JS (CEF-backed) — replaces the old Slate window.
- Added Anthropic (Claude), DeepSeek, and Custom (OpenAI-compatible) providers — old build only had Gemini + OpenAI.
- Added 5-slot API key storage (was: single key).
- Removed the built-in trial key and the per-token usage cap — bring your own key from first launch.
- Added image-paste support (Ctrl+V).
- Modernised context extraction (collapsed graphs, macros, performance scan, complexity score).
- Added themes (preset + custom accent).
- Conversation storage moved to a versioned JSON format; chats from 1.x are read-as-is (no migration needed).
v1.0
- Initial release with Gemini + OpenAI cloud providers, single API key, basic Slate UI.
Thank you for using Blueprint Analyst!