ldraw-mcp

ldraw-mcp

Renders LEGO LDraw model files (.ldr, .mpd, .dat) to high-quality images using headless Blender, enabling MCP clients to visually inspect LEGO builds.

Category
Visit Server

README

<div align="center">

ldraw-mcp

Give your MCP client eyes for LEGO® models.

Render LDraw files (.ldr / .mpd / .dat) to images with real part geometry — studs, slopes, window glass — using headless Blender and the ImportLDraw addon. The output looks like a BrickLink Stud.io render, with no GUI anywhere in the loop.

PyPI Python License: MIT MCP CI

<img src="docs/hero.png" alt="Two rendered views of a red LEGO car — real studs, transparent glass, rubber tires, steering wheel" width="100%">

<sub>A ~90-line <code>.ldr</code> rendered front-left and rear-right — actual bricks, not a geometric proxy.</sub>

</div>

<!-- mcp-name: io.github.musharna/ldraw-mcp -->


Point a vision-capable model at a build and it sees the actual bricks: crossed rotation matrices, floating plates, sunken windows — the kinds of export bugs a geometric proxy render will happily hide.

Quickstart

# 1. install
pip install ldraw-mcp

# 2. install the LDraw parts library + ImportLDraw addon
ldraw-mcp-setup

# 3. register with Claude Code
claude mcp add ldraw -- ldraw-mcp

Then ask things like "render output/build.ldr and tell me what looks wrong" — the model sees the render, not just the text.

Blender is a prerequisite (see Requirements); it is not installed by ldraw-mcp-setup.

Tools

tool what it does
render_ldraw_file(path, azimuths="-60,120", resolution=640, samples=24) Render a model file to a PNG (multi-view, stitched side by side)
render_ldraw_text(ldr, azimuths="-60,120", resolution=640, samples=24) Render inline LDraw content without writing a file first
check_renderer() Diagnose the Blender / addon / parts-library setup

azimuths is a comma-separated list of view angles in degrees; each is rendered and the views are stitched horizontally. Elevation is fixed at 22°. Higher samples = cleaner but slower.

Requirements

  • Blender 4.x on PATH, or point LDRAW_MCP_BLENDER at the binary. Install it yourself (package manager, blender.org, or a local build); ldraw-mcp-setup does not install Blender.
  • ImportLDraw addon (io_scene_importldraw) in Blender's addons dir — installed by ldraw-mcp-setup.
  • LDraw parts library at ~/.ldraw (or LDRAW_LIBRARY_PATH) — installed by ldraw-mcp-setup.

Environment variables

var meaning
LDRAW_MCP_BLENDER Path to the blender binary (overrides PATH lookup)
LDRAW_MCP_DISABLE Set to 1 to force is_available() to False
LDRAW_LIBRARY_PATH Path to the LDraw parts library (community convention)

Manual setup

If ldraw-mcp-setup can't detect things automatically:

  • LDraw library: download complete.zip and unzip so that ~/.ldraw/parts/ exists.
  • ImportLDraw addon: download the latest release from TobyLobster/ImportLDraw and install it via Blender > Preferences > Add-ons > Install, or unzip into ~/.config/blender/<version>/scripts/addons/io_scene_importldraw/. (Launch Blender once first so the config directory exists.)

Troubleshooting

  • check_renderer says NOT FOUND: run ldraw-mcp-setup, or set the relevant env var above.
  • No GPU / WSL2 / containers: rendering uses Cycles on CPU, which works headless everywhere — no GPU or display needed. A ~150-part model takes a few seconds at the default 640px / 24 samples.
  • "no mesh objects imported": the addon couldn't resolve parts — usually a wrong or incomplete LDraw library path. Re-run setup or check LDRAW_LIBRARY_PATH.
  • Addon not enabled: the render script enables it automatically per run; if a manual Blender session complains, enable io_scene_importldraw in Preferences > Add-ons.

Provenance

This renderer was extracted from the prompt2brick project, where it started life as the vision critic's "see the actual model" path. prompt2brick keeps its own vendored copy of the render wrapper and Blender script, but this repo is the canonical source going forward — fixes and improvements to the renderer should land here first and be ported back into prompt2brick.

License

MIT — see LICENSE.


<sub>LEGO® is a trademark of the LEGO Group, which does not sponsor, authorize, or endorse this project. This tool is not affiliated with the LEGO Group, BrickLink, or the LDraw.org organization.</sub>

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
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
E2B

E2B

Using MCP to run code via e2b.

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
Qdrant Server

Qdrant Server

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

Official
Featured