MCP Server

The MCP Server is a bridge between NodeLens and Claude Desktop. Once connected, Claude can read your entire project — every Blueprint, every function, every variable, and every note — and help you analyze, document, and generate graph logic without you having to copy and paste anything manually.

What is MCP?

MCP (Model Context Protocol) is an open standard developed by Anthropic that allows AI assistants to connect to external tools and data sources. NodeLens includes a built-in MCP server that Claude Desktop can connect to directly.

Think of it as giving Claude a window into your NodeLens workspace — it can browse your projects, open any function, read its logic, and save results back to disk, all within a normal conversation.

Setup

You will need Claude Desktop installed and Node.js on your machine. Node.js is free and available at nodejs.org — download the LTS version if you are unsure which one to pick.

1. Find the Claude Desktop config file.
   Open File Explorer and navigate to:

   %APPDATA%\Claude\

   You are looking for a file called claude_desktop_config.json. If it does not exist, create it — right-click in the folder, select New → Text Document, rename it to claude_desktop_config.json (make sure the .txt extension is removed).
2. Open the file in a text editor.
   Right-click the file and open it with Notepad or any text editor.
3. Add the NodeLens MCP server configuration.
   Paste the following into the file (replace the path with your actual NodeLens installation path):

   {
     "mcpServers": {
       "nodelens": {
         "command": "node",
         "args": ["C:\\Users\\YourName\\AppData\\Local\\NodeLens\\mcp-server\\index.js"]
       }
     }
   }

   The mcp-server folder is located inside the NodeLens installation directory, typically at C:\Users\YourName\AppData\Local\NodeLens\.
4. Save the file and restart Claude Desktop.
   Close Claude Desktop completely and reopen it. After restarting, open a new conversation and look for the NodeLens icon in the tools area — this confirms the connection is active.

What You Can Ask Claude to Do

Once connected, you can talk to Claude in plain language. Here are some things you can ask:

• Browse your projects "What projects do I have in NodeLens?" — Claude lists all your projects and Blueprints.
• Read and explain a function "Read the TakeDamage function in BP_Character and explain what it does." — Claude reads the pseudocode and gives you a plain-language explanation.
• Find a specific function "Search for any function that deals with inventory pickup." — Claude searches across all your Blueprints and returns matching functions.
• Understand relationships between Blueprints "How does BP_PlayerController communicate with BP_HUD?" — Claude reads multiple Blueprints and builds a picture of how they connect.
• Generate a new function "Create a new function called ReloadWeapon that checks ammo and plays a montage." — Claude proposes the logic, confirms with you, then saves it as a new Blueprint file ready to import into UE5.
• Modify an existing function "Add a health check to the Sprint function — if health is below 20%, sprinting should be disabled." — Claude reads the existing function, proposes the change, and saves a modified version to a new folder.
• Generate documentation "Create a markdown file that documents all functions in BP_InventoryComponent with a description and parameter list." — Claude reads each function and saves a structured MD file directly into NodeLens.
• Answer architecture questions "Is there any Blueprint that directly sets the player's health without going through a damage function?" — Claude scans your project and reports what it finds.

Best Practices

• ✓
  Always save AI changes to a new folder. When you re-send a Blueprint from Unreal Engine, NodeLens overwrites the existing data. Any AI-generated changes saved to the original Blueprint files will be lost. Ask Claude to save results to a separate folder (e.g. "Fixes" or "AI Generated") so they are never overwritten.
• ✓
  Start with "read" before "modify". Ask Claude to read and explain a function before asking it to change it. This gives Claude full context and reduces the chance of errors.
• ✓
  Confirm pseudocode before generating graph logic. When asking Claude to create or modify a function, ask it to show you the pseudocode first. Approve the logic before Claude writes the graph — it is much easier to catch mistakes at this stage.
• ✓
  Use Custom Instructions for project context. Add key information about your project to Custom Instructions in Settings — struct definitions, naming conventions, architecture rules. Claude will apply this context automatically in every session.
• ✓
  Always review the Canvas after importing AI-generated graphs. Open the result in the Canvas tab to visually verify connections and logic before bringing it into your UE5 project.