AutoCAD Electrical MCP (ECM)

MCP server + local AI web dashboard for AutoCAD Electrical 2027 - control AutoCAD with plain language via Claude, Ollama, or any OpenAI compatible models.

This fork was designed for Electrical Controls Manufacturing (ECM), aimed at adding support for their Motor Control Center (MCC) products, as well as planned support for variable frequency drives (VFDs), soft starters, breaker panels, and other products. The MCC configurator will not work unless MCC_LAYOUT, MCC_UNITDATA, and MCC_NAMEPLATE.dwg are open in AutoCAD, and the required CAD blocks are located in the filepath written in .env. The original repo and its README.md are available at https://github.com/Igualguana/AUTOCAD-ELECTRICAL-MCP.

MCC Configurator Setup

The MCC Builder requires three template drawings to be open in AutoCAD before you start a project. These are the files the tool writes into — they must come from your company's block library.

Steps:

1. Set MCC_BLOCK_LIBRARY in your .env to the folder containing these files (and all unit block DWGs).

2. Open MCC_LAYOUT.dwg, MCC_UNITDATA.dwg, and MCC_NAMEPLATE.dwg in AutoCAD Electrical.

3. Start the web server: python start_web.py

4. Open http://127.0.0.1:8080 and click MCC in the sidebar.

5. Click New Project — the tool will connect to the three open drawings automatically.

If you see "MCC_LAYOUT.dwg is not open" when creating a project, go back to step 2 and make sure all three files are open in AutoCAD before clicking New Project.

Related MCP server: civil3d-mcp

Web Interface

The web dashboard is a FastAPI + vanilla JS single-page app served locally at http://127.0.0.1:8080.

Running locally

# Start the web server (opens browser automatically)
python start_web.py

# Or specify host/port explicitly
python start_web.py --host 127.0.0.1 --port 8080 --no-browser

Requirements

• Windows 10/11 (64-bit) — COM automation is Windows-only

• Python 3.11+

• AutoCAD Electrical 2027 running with the required .dwg files open

• Mode A only: Anthropic API key

• Mode B only: Ollama installed and running

Installation

Option A — One-click installer (recommended)

1. Download or clone the repo — if you don't have Git yet, download the ZIP from GitHub and unzip it.

2. Open the folder, right-click install.bat → Run as administrator.

The script will:

• Install Python 3.11 via winget if it isn't already on your machine

• Install Git via winget if it isn't already on your machine

• Install all required Python libraries automatically

• Copy .env.example → .env so you can fill in your settings

winget is built into Windows 10 (1709+) and Windows 11. If your machine doesn't have it, install Python 3.11+ and Git manually, then run Option B below.

Option B — Manual install

# 1. Clone
git clone https://github.com/nihaalgaina/AUTOCAD-ELECTRICAL-MCP
cd AUTOCAD-ELECTRICAL-MCP

# 2. Install dependencies
python scripts/install.py

Configure your environment

Open .env in any text editor and fill in the values for your setup:

# Path to the folder containing your MCC block .DWG files
# (MCC_400.DWG, UNIT4040.DWG, UDATALIN.DWG, LAMACOID.DWG, etc.)
MCC_BLOCK_LIBRARY=C:\Path\To\Your\MCC Blocks

# Only needed if you want to use Claude as the AI provider
ANTHROPIC_API_KEY=sk-ant-...

# Only needed for OpenAI / Groq
OPENAI_API_KEY=...
GROQ_API_KEY=...

MCC_BLOCK_LIBRARY is the only setting most users need to change. Point it at whichever folder holds your company's .DWG block library. Multiple folders are supported — separate them with a semicolon: MCC_BLOCK_LIBRARY=C:\Blocks\400mm;C:\Blocks\500mm

Ollama setup (Mode B — local AI, no API key required)

# Install from https://ollama.com then:
ollama serve                   # start the server
ollama pull qwen2.5:0.5b       # fast, runs on 4 GB RAM
# or
ollama pull qwen2.5:7b         # better quality, needs ~6 GB RAM

Recommended: qwen2.5:0.5b (397 MB) works well for the keyword pre-router and is reliable on machines with limited RAM. For the full LLM path use qwen2.5:7b or larger.

Author (Original Repository)

Randy Igualguana
Copyright © 2026 Randy Igualguana

Built with Model Context Protocol · FastMCP · FastAPI · pywin32 · Ollama