ShapeItUp
Enables AI agents to write TypeScript to create, render, verify, and export parametric 3D models using Replicad/OpenCascade CAD kernel, headlessly from terminal or MCP clients.
README
<p align="center"> <img src="logo.png" alt="ShapeItUp" width="400"> </p>
<p align="center"> <strong>Scripted CAD for AI agents and humans</strong> — write TypeScript, get CAD. </p>
<p align="center"> <a href="https://github.com/asbis/ShapeItUp/blob/master/LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue.svg" alt="License: MIT"></a> <a href="https://www.npmjs.com/package/@shapeitup/mcp-server"><img src="https://img.shields.io/npm/v/@shapeitup/mcp-server.svg" alt="npm: @shapeitup/mcp-server"></a> </p>
ShapeItUp is an MCP server that turns Replicad / OpenCascade into AI-agent-grade CAD tooling. It writes, renders, verifies, and exports parametric 3D models from TypeScript .shape.ts files, headlessly — from the terminal, from Claude Code / Cursor / Claude Desktop, or from CI. A VSCode extension is available as an optional interactive viewer for humans who want to watch the renders happen live.
Install for Claude Code
claude mcp add shapeitup -- npx -y @shapeitup/mcp-server
That's it. claude mcp add registers the server in your Claude Code config so agent sessions pick it up automatically. See the Claude Code MCP docs for scoping options (-s user for all projects, -s project for just this repo).
Install for Cursor / Claude Desktop / other MCP clients
The generic MCP-config snippet:
{
"mcpServers": {
"shapeitup": {
"command": "npx",
"args": ["-y", "@shapeitup/mcp-server"]
}
}
}
Where to put it:
- Cursor —
~/.cursor/mcp.json, or Settings → MCP → Add new server to drop it in through the UI. - Claude Desktop —
~/Library/Application Support/Claude/claude_desktop_config.jsonon macOS,%APPDATA%\Claude\claude_desktop_config.jsonon Windows. Restart the app after editing. - Any stdio MCP client — hand it
npx -y @shapeitup/mcp-serveras the server command.
Node 20+ required. No native build step; everything ships as WASM.
What you get
A CAD toolkit agents can drive end to end:
- Authoring —
create_shape,modify_shape,read_shape,list_shapes,setup_shape_project - Rendering —
render_preview,preview_shape,get_preview(SVG-first, resvg-backed PNGs) - Verification —
verify_shape,check_collisions,sweep_check,describe_geometry,validate_joints - Iteration —
tune_params(slider overrides),set_render_mode,toggle_dimensions - Export —
export_shape(STEP / STL / OBJ / 3MF) - stdlib —
holes,screws/bolts/washers/inserts,bearings,extrusions,patterns,threads,joints+assemble,printHints, and more — all importable from"shapeitup"inside any.shape.ts
Full tool list: 25 MCP tools covering the create → preview → verify → tune → export loop.
Quick example
Prompt to the agent: "I want a mounting bracket, 60×40×4 mm, four M3 through-holes in the corners." The agent calls create_shape and writes:
// bracket.shape.ts
import { drawRoundedRectangle } from "replicad";
import { holes } from "shapeitup";
export const params = { width: 60, depth: 40, thickness: 4, holeInset: 6 };
export default function main({ width, depth, thickness, holeInset }: typeof params) {
const plate = drawRoundedRectangle(width, depth, 3)
.sketchOnPlane("XY")
.extrude(thickness);
return holes.through(plate, "M3", [
[-width/2 + holeInset, -depth/2 + holeInset, 0],
[ width/2 - holeInset, -depth/2 + holeInset, 0],
[-width/2 + holeInset, depth/2 - holeInset, 0],
[ width/2 - holeInset, depth/2 - holeInset, 0],
]);
}
Then it calls render_preview and gets back a PNG it can inspect, plus bounding-box + volume metadata to sanity-check against the spec. If the user wants the holes bigger, the agent calls tune_params — no file rewrite needed.
Running in CI / containers
npx -y @shapeitup/mcp-server works anywhere Node 20 runs — Docker, GitHub Actions, GitLab CI. Renders go through @resvg/resvg-wasm; no GPU, no headless Chromium, no native graphics stack required. A cold boot on a stock Node image typically takes ~3 s for the first render (OCCT WASM compile) and < 200 ms thereafter.
VSCode extension (optional)
There is a VSCode extension that subscribes to the MCP server's event bus and shows a live Three.js viewer in the side panel — useful when you want to watch an agent work, or to drive ShapeItUp interactively. Install it from the VS Code marketplace (search for "ShapeItUp") or build from source (see Development).
The extension is entirely optional: the MCP server is headless on its own, does not require VSCode to be running, and produces the same renders whether or not a viewer is attached.
Development
git clone https://github.com/asbis/ShapeItUp.git
cd ShapeItUp
pnpm install
pnpm build # builds extension, viewer, worker, and mcp-server
pnpm dev # watch mode
pnpm lint # tsc -b across packages
Then press F5 in VS Code to launch the Extension Development Host if you want to iterate on the viewer.
Project layout:
packages/
mcp-server/ -- standalone MCP server (this is the npm package)
core/ -- Replicad wrappers, stdlib, execution engine (bundled into mcp-server)
shared/ -- cross-package types (bundled into mcp-server)
extension/ -- VSCode extension (optional viewer)
viewer/ -- Three.js viewer surface for the extension
worker/ -- OCCT WASM web-worker for the extension viewer
See CONTRIBUTING.md for contribution guidelines.
License
MIT
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
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.