MCP Revit Integration

Overview

This project enables AI-powered interaction with Autodesk Revit through the Model Context Protocol (MCP). It combines:

• A Node.js MCP server that exposes Revit operations as AI-accessible tools.
• A Revit plugin that executes commands, queries model data, and communicates with the MCP server.

Together, they allow AI clients (like Cline or Claude) to analyze, modify, and automate Revit models.

---

Features

• Query Revit model data (views, elements, families, parameters)
• Create, modify, and delete Revit elements (walls, doors, floors, etc.)
• Color-code elements based on parameters
• Tag elements automatically
• Execute custom C# code snippets inside Revit
• Extendable with your own Revit commands and MCP tools

---

Architecture

flowchart LR
    subgraph AI_Client
        Cline
    end
    subgraph MCP_Server
        revit-mcp
    end
    subgraph Revit
        subgraph Plugin
            SocketService
            CommandManager
            CommandSet
            CommandExecutor
        end
    end

    Cline --> revit-mcp
    revit-mcp --> SocketService
    SocketService --> CommandManager
    CommandManager --> CommandSet
    CommandSet --> CommandExecutor
    CommandExecutor --> SocketService

---

Components

1. MCP Server (revit-mcp)

• Node.js app exposing Revit operations as MCP tools
• Communicates with AI clients
• Sends commands to the Revit plugin via socket

2. Revit Plugin (revit-mcp-plugin)

• .NET plugin loaded into Revit (2019-2024)
• Registers and executes commands
• Communicates with MCP server
• Loads Command Sets (DLLs) for specific features

---

Installation

Prerequisites

• Node.js 18+
• Autodesk Revit 2019-2024
• Visual Studio (to build the plugin)

1. Build MCP Server

npm install
npm run build

2. Build Revit Plugin

• Open plugin/revit-mcp-plugin.sln in Visual Studio
• Set configuration to Release and platform to x64
• Build the solution
• Copy revit-mcp-plugin.dll and revit-mcp-sdk.dll to:

C:\Users\<YourUser>\AppData\Roaming\Autodesk\Revit\Addins\2022\

3. Register the Plugin

Create a .addin file in the same folder:

<?xml version="1.0" encoding="utf-8"?>
<RevitAddIns>
  <AddIn Type="Application">
    <Name>revit-mcp</Name>
    <Assembly>C:\Users\<YourUser>\AppData\Roaming\Autodesk\Revit\Addins\2022\revit-mcp-plugin.dll</Assembly>
    <FullClassName>revit_mcp_plugin.Core.Application</FullClassName>
    <ClientId>090A4C8C-61DC-426D-87DF-E4BAE0F80EC1</ClientId>
    <VendorId>revit-mcp</VendorId>
    <VendorDescription>https://github.com/BTankut/MCP_Revit</VendorDescription>
  </AddIn>
</RevitAddIns>

4. Configure MCP Server in Cline

Add to your cline_mcp_settings.json:

{
  "mcpServers": {
    "revit-mcp": {
      "command": "node",
      "args": ["c:/Users/BT/CascadeProjects/MCP Revit/build/index.js"],
      "env": {}
    }
  }
}

---

Usage

• Start Revit. The plugin loads automatically.
• Start the MCP server:

node build/index.js

• Open Cline or Claude.
• Use AI commands like:
  • "List all walls"
  • "Create a new door here"
  • "Color rooms by area"
  • "Delete selected elements"

---

Extending Functionality

Add New Revit Commands

• Develop C# commands in a Command Set (DLL)
• Register them in command.json
• Place in Commands/YourCommandSet/2022/YourCommandSet.dll

Add New MCP Tools

• Create new .ts files in src/tools/
• Implement tool logic (call plugin commands, process data)
• Export as MCP tool
• Rebuild server (npm run build)

---

Project Structure

• src/ — MCP server TypeScript source
• build/ — Compiled MCP server
• plugin/ — Revit plugin source
• plugin/bin/x64/Release/ — Compiled plugin DLLs
• plugin/SampleCommandSet/ — Example command set
• sdk_extract/ — Extracted SDK dependencies

---

Summary

This project enables AI-driven BIM automation by bridging Revit and AI clients via MCP:

• AI issues commands
• MCP server translates and forwards
• Revit plugin executes and responds

You can customize both plugin commands and MCP tools to fit your workflows.

---

License

MIT License