DocsGetting Started

Getting Started

This guide walks you through installing the plugin, setting up an AI agent, and running your first prompt.

# Prerequisites

- At least one AI agent CLI installed, or an OpenRouter API key for the built-in agent.

- Node.js 18+ — Required for CLI agents (Claude Code, Gemini CLI, Codex, etc.). The plugin auto-detects Node.js from common install paths and version managers (nvm, volta, fnm, bun, mise).

- Unreal Engine 5.5, 5.6, or 5.7 (source or Launcher build)

# Installation

From Fab

  1. 1. Purchase or download Blueprint AI from the Unreal Marketplace (Fab).
  2. 2. Enable the plugin in Edit > Plugins and restart the editor.

Manual Install

  1. 1. Copy the BlueprintAI folder into your project's Plugins/ directory.
  2. 2. Regenerate project files and restart the editor.

# Opening the Chat Panel

Once the plugin is enabled, open the chat panel from the menu bar:

Window > AI > Blueprint AI

The panel opens as a dockable tab. You can have multiple chat tabs open simultaneously, each connected to a different agent or session.

# Agent Setup

Blueprint AI supports two categories of agents:

Built-in OpenRouter Agent (no CLI required)

  1. 1. Go to Project Settings > Plugins > Blueprint AI.
  2. 2. Paste your OpenRouter API key in the OpenRouter API Key field. Get one at openrouter.ai.
  3. 3. Optionally change the Default OpenRouter Model (defaults to anthropic/claude-sonnet-4). You can also switch models per session from the chat UI.

CLI Agents (Claude Code, Gemini CLI, Codex, etc.)

CLI agents run as external processes and communicate via JSON-RPC over stdio. In the agent picker, each runtime also shows a status label such as Installed, Not installed, or Needs API key. Blueprint AI does not auto-download CLIs or ACP adapters anymore. Install required runtimes and ACP adapters manually in your terminal:

# Claude Code
curl -fsSL https://claude.ai/install.sh | bash
npm install -g @zed-industries/claude-code-acp
# Codex CLI
npm install -g @openai/codex
npm install -g @zed-industries/codex-acp
# Gemini / Copilot / OpenCode
npm install -g @google/gemini-cli
npm install -g @github/copilot
npm i -g opencode-ai

Some agents require their own authentication (e.g., running claude login in your terminal before first use).

Agent Picker Behavior

- Selecting a different agent from the runtime picker opens a new tab for that runtime.

- Your current chat tab is preserved, so you can keep multiple concurrent sessions (for example, one Claude tab and one Codex tab).

# Your First Prompt

  1. 1. Select an agent from the dropdown.
  2. 2. Type a message in the input field, for example:
Read the Blueprint at /Game/ThirdPerson/BP_ThirdPersonCharacter and summarize its event graph.
  1. 3. The agent will call the read_file tool, read the Blueprint's structure, and respond with a summary.

# Attaching Context

You can provide additional context to the AI by attaching Blueprint nodes, assets, or images:

- Right-click a node in the Blueprint graph editor and select Select as Context. The node's pins, connections, and properties are serialized and sent with your next prompt.

- Right-click an asset in the Content Browser and select Select as AI Context. The Blueprint's structure overview (variables, components, functions, events) is serialized.

- Drag and drop Blueprint assets directly onto the chat input area.

- Paste an image from your clipboard (Ctrl/Cmd+V) into the chat input. The image is base64-encoded and sent to vision-capable models.

# What's Next?

- Explore the full list of 25+ editor tools in the Tools Reference.

- If anything fails during setup, use the Troubleshooting page for quick fixes (Node path issues, expired OAuth tokens, missing CLIs, and more).

- Connect external MCP clients to the embedded MCP server for programmatic access.