Greenloom CAD MCP Server

Greenloom CAD MCP Server

Enables automated CAD operations via natural language, supporting both AutoCAD LT on Windows and headless DXF generation on any platform.

Category
Visit Server

README

Greenloom CAD MCP Server

MCP server for AutoCAD LT automation and headless DXF generation.

Two backends, one API:

Backend Runtime Requires AutoCAD? Screenshot
File IPC Windows Python Yes — AutoCAD LT 2024+ (Windows) Win32 PrintWindow
ezdxf Any platform No (headless) matplotlib render

The server exposes 8 consolidated tools (drawing, entity, layer, block, annotation, pid, view, system) over the MCP stdio transport. An MCP client (Claude Desktop, Claude Code, etc.) connects and drives AutoCAD through natural-language requests.

Prerequisites (File IPC backend)

  • Windows 10/11 (the File IPC backend uses Win32 APIs for focus-free window messaging)
  • AutoCAD LT 2024 or newer — AutoLISP support was added in LT 2024 for Windows. AutoCAD LT for Mac exists but does not support AutoLISP.
  • Python 3.10+ (Windows native — not WSL Python)
  • uv package manager (install guide)

The ezdxf headless backend works on any platform (Linux, macOS, WSL) for offline DXF generation without AutoCAD installed.

Quick Start

1. Clone and install

git clone https://github.com/Greenmint-labs/greenloom_CAD_MCP.git
cd greenloom_CAD_MCP
uv sync

2. Load the LISP dispatcher in AutoCAD LT

Open AutoCAD LT and load mcp_dispatch.lsp using APPLOAD:

  1. Type APPLOAD in the AutoCAD command line
  2. Browse to <repo>/lisp-code/mcp_dispatch.lsp
  3. Click Load
  4. You should see: === MCP Dispatch v3.1 loaded === and Ready for commands via (c:mcp-dispatch)

Tip: Add the file to your AutoCAD Startup Suite (in the APPLOAD dialog) so it loads automatically with every drawing.

3. Configure your MCP client

Add to your MCP client configuration (e.g. Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "greenloom-cad-mcp": {
      "command": "C:\\path\\to\\greenloom-cad-mcp\\.venv\\Scripts\\python.exe",
      "args": ["-m", "greenloom_cad_mcp"],
      "env": { "GREENLOOM_CAD_BACKEND": "auto" }
    }
  }
}

Key points:

  • The command must point to the Windows Python inside the project venv (not WSL python).
  • GREENLOOM_CAD_BACKEND can be auto (default — tries File IPC, falls back to ezdxf), file_ipc (requires AutoCAD), or ezdxf (headless only).

Running from WSL

If your MCP client runs in WSL (e.g. Claude Code), launch the server through cmd.exe so it runs as a native Windows process:

{
  "mcpServers": {
    "greenloom-cad-mcp": {
      "type": "stdio",
      "command": "cmd.exe",
      "args": ["/d", "/s", "/c", "cd /d C:\\path\\to\\greenloom-cad-mcp && .venv\\Scripts\\python.exe -m greenloom_cad_mcp"],
      "env": { "GREENLOOM_CAD_BACKEND": "auto" }
    }
  }
}

4. Verify

From your MCP client, call:

system(operation="status")

You should see backend: "file_ipc" if AutoCAD is running, or backend: "ezdxf" for headless mode.

Tools

drawing — File/drawing management

Operation Description File IPC ezdxf
create Reset to clean drawing (erase all + purge) Yes Yes
open Open an existing drawing Yes Yes (DXF)
info Get entity count and layers Yes Yes
save Save current drawing (to path if given) Yes Yes
save_as_dxf Export as DXF Yes Yes
plot_pdf Plot to PDF Yes No
purge Purge unused objects Yes Yes
get_variables Get system variables by name Yes Yes
undo Undo last operation Yes No
redo Redo last undone operation Yes No

entity — Entity CRUD + modification

Create: create_line, create_circle, create_polyline, create_rectangle, create_arc, create_ellipse, create_mtext, create_hatch

Read: list, count, get

Modify: copy, move, rotate, scale, mirror, offset*, array, fillet*, chamfer*, erase

* offset, fillet, chamfer are File IPC only (not supported in ezdxf headless backend).

layer — Layer management

list, create, set_current, set_properties, freeze, thaw, lock, unlock

block — Block operations

Operation File IPC ezdxf
list Yes Yes
insert Yes Yes
insert_with_attributes Yes Yes
get_attributes Yes Yes
update_attribute Yes Yes
define No Yes

annotation — Text, dimensions, leaders

create_text, create_dimension_linear, create_dimension_aligned, create_dimension_angular, create_dimension_radius, create_leader

pid — P&ID operations (CTO symbol library)

setup_layers, insert_symbol, list_symbols, draw_process_line, connect_equipment, add_flow_arrow, add_equipment_tag, add_line_number, insert_valve, insert_instrument, insert_pump, insert_tank

P&ID symbol insertion requires the CAD Tools Online (CTO) P&ID Symbol Library installed at C:\PIDv4-CTO\. The ezdxf backend has built-in CTO library support. For the File IPC backend, some P&ID operations require additional LISP helpers — see the P&ID section in the wiki for setup details.

view — Viewport and screenshot

Operation Description
zoom_extents Zoom to show all entities
zoom_window Zoom to a specified window
get_screenshot Capture current AutoCAD view as PNG

Screenshots use PrintWindow (Win32) for the File IPC backend — works even when AutoCAD is minimized or in the background. The ezdxf backend renders via matplotlib.

system — Server management

status, health, get_backend, runtime, init, execute_lisp

execute_lisp runs arbitrary AutoLISP code (File IPC only). Pass data: {code: "(+ 1 2)"}. This turns the server into an extensible automation platform — any valid AutoLISP expression can be executed.

Architecture

MCP Client (Claude)
    │  stdio (JSON-RPC)
    ▼
Python MCP Server (greenloom_cad_mcp)
    │
    ├── File IPC Backend ──► C:/temp/*.json ──► mcp_dispatch.lsp (AutoCAD LT)
    │   PostMessageW(WM_CHAR) to MDIClient — no focus steal
    │
    └── ezdxf Backend ──► in-memory DXF (headless, no AutoCAD needed)

The File IPC backend sends keystrokes to AutoCAD's MDIClient window via PostMessageW(WM_CHAR), triggering the (c:mcp-dispatch) AutoLISP command. This approach does not steal window focus — you can continue working in other applications while automation runs.

Environment Variables

Variable Default Description
GREENLOOM_CAD_BACKEND auto Backend selection: auto, file_ipc, ezdxf
GREENLOOM_CAD_IPC_DIR C:/temp Directory for IPC command/result JSON files (must match on both Python and LISP sides)
GREENLOOM_CAD_IPC_TIMEOUT 10.0 IPC command timeout in seconds (1-300)
GREENLOOM_CAD_ONLY_TEXT false Disable screenshot capture (text feedback only)

Note: If you change GREENLOOM_CAD_IPC_DIR, you must also update the *mcp-ipc-dir* variable in mcp_dispatch.lsp to match.

Development

uv sync
uv run pytest tests/ -v

AutoCAD LT AutoLISP Compatibility

AutoLISP was added to AutoCAD LT in the 2024 release (Windows only). AutoCAD LT for Mac does not support AutoLISP.

Supported (LT 2024+ Windows) Not Supported
.lsp / .fas / .vlx / .dcl VLIDE (Visual LISP IDE)
All vl-* utility functions vlax-* (ActiveX/COM)
File I/O (open, read-line, etc.) Express Tools
Entity access (entget, entmod, etc.) 3D operations
Selection sets AutoLISP on Mac

The mcp_dispatch.lsp dispatcher is fully compatible with LT 2024+.

What's New in v3.1

  • execute_lisp — Run arbitrary AutoLISP code via temp file pattern. Turns the server from a fixed command set into an extensible automation platform.
  • Undo / Redo — Single-step undo and redo via drawing tool.
  • Drawing open — Open existing .dwg files programmatically (FILEDIA suppressed).
  • Drawing create — Now resets current drawing (erase all + purge) instead of _.NEW, preserving the LISP dispatcher namespace.
  • Drawing save with pathsave with a path parameter uses SAVEAS; without path uses QSAVE.
  • get_variables fix — Respects the names parameter; returns requested variables with proper type handling.
  • Polyline/leader fix — Point arrays properly encoded via semicolon-delimited format.
  • ESC prefix — Sends 2x ESC before each dispatch to cancel stale pending commands from prior timeouts.
  • UTF-8/cp1252 fallback — Handles non-ASCII characters in LISP result files (AutoCAD writes Windows-1252).
  • Configurable IPC timeoutGREENLOOM_CAD_IPC_TIMEOUT env var (1–300 seconds, default 10).
  • Thread-safe backend initasyncio.Lock prevents parallel initialization races.

License

MIT

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured