Click ○ to set current · double-click a name to rename

Cur	Name	On	Freeze	Lock	Color	Linetype	Lineweight	Objects

🤖

JsCAD Generative CAD MCP Server & Browser Bridge

Empowers AI Agents (Antigravity, Claude, Cursor, Windsurf, BrowserOS) to programmatically compose, calculate, and stream massive 2D CAD engineering drawings with 0 token exhaustion.

🚀 Why MCP & Generative CAD?

Traditional AI interactions with CAD suffer from a severe limitation: outputting 10,000 raw drawing coordinates via text prompts exhausts LLM context windows, costs hundreds of thousands of tokens, and is painfully slow. The JsCAD CAD Engine MCP Server eliminates this bottleneck by giving AI models programmatic CAD drafting tools and an isolated JavaScript execution sandbox.

⚡ Key Superpowers for AI Agents

Procedural Code Sandbox (cad_execute_script): AI writes high-speed JavaScript generator loops. 50,000 entities are calculated and rendered in milliseconds inside an isolated Node permission-checked process.
High-Level Parametric Generators (cad_generate_parametric): Built-in instant generators for structural column grids, roof trusses, machine flanges with bolt circles, involute gear teeth, and multi-room floor plans.
Real AutoCAD Command Pipeline (cad_run_commands): Executes authentic AutoCAD command strings (e.g. LINE 0,0 100,0; CHPROP ALL; LA WALLS C red) with Quick Select filters and full problem reporting.
Bi-Directional DXF & SVG Plotting: Imports and exports full AutoCAD 2000 (AC1015) ASCII DXF with layers, linetypes, colors, and blocks, plus instant vector SVG visual verification.
Cloud Collaboration (cad_share_drawing): Shares active sessions to Cloudflare D1 with live view & edit URLs for real-time human-agent co-designing.

🏗️ System Architecture

+--------------------------+         +-------------------------------+
| AI Agents (Claude /      |         | Browser CAD Workspace (JsCAD) |
| Antigravity / Cursor)    |         | https://autocad.site          |
+--------------------------+         +-------------------------------+
       |                                             ^
       | stdio JSON-RPC                              | Live Hash / WebSocket / BroadcastChannel
       v                                             v
+--------------------------+  Cloudflare D1  +-------------------------------+
| CAD Engine MCP Server    | <=============> | CAD Web Bridge (window.CAD)   |
| (mcp-server/index.js)    |   Share & Sync  | URL params (?cmd=, ?sample=)  |
+--------------------------+                 +-------------------------------+
       |
       +---> Node Permission Sandbox (cad_execute_script)
       +---> Parametric Geometry Generators (Truss, Grid, Flange, Gear)
       +---> AC1015 DXF Importer & Exporter / SVG Vectorizer

🛠️ Complete MCP Tool Inventory (16 Tools)

Every tool is available over JSON-RPC via stdio to any MCP-compliant client:

Tool Name	Description	Key Parameters
`cad_new_drawing`	Initializes or resets drawing space, filename, and engineering units.	`name`, `units`
`cad_add_layers`	Creates layers with standard AutoCAD ACI colors, linetypes, and lineweights.	`layers: [{ name, color, linetype, lineweight }]`
`cad_batch_add_entities`	High-speed bulk insertion of primitives without command line overhead.	`entities: [ { type, ...props } ]`
`cad_generate_parametric`	Algorithmic generators for complex engineering assemblies.	`type`: `structural_grid`, `truss`, `mechanical_flange`, `gear`, `architectural_floorplan`
`cad_run_commands`	Runs native AutoCAD command-line strings with optional Quick Select filtering.	`commands: "LINE 0,0 100,0; ZOOM E"`, `select: {...}`
`cad_list_commands`	Searches and lists all implemented CAD commands, aliases, and descriptions.	`filter: "layer"`
`cad_execute_script`	Runs procedural JavaScript generator loops in an isolated Node sandbox.	`code: "for(let i=0; i<100; i++) ..."`
`cad_create_block`	Defines reusable Blocks with geometry and dynamic `ATTDEF` tags.	`name`, `basePoint`, `entities`
`cad_insert_block`	Inserts block references with position, scale, rotation, and attribute values.	`name`, `position`, `scale`, `rotation`, `attributes`
`cad_open_dxf`	Opens an existing ASCII DXF (AutoCAD R12 to 2018+) into the active session.	`filename: "drawing.dxf"`
`cad_export_dxf`	Exports AutoCAD 2000 (AC1015) compliant ASCII DXF with symbol tables.	`filename: "output.dxf"`
`cad_export_svg`	Renders high-fidelity vector SVG preview for visual agent inspection.	`filename: "preview.svg"`, `width`, `height`
`cad_get_stats`	Inspects drawing metrics: entity count, layer distribution, and bounds.	None
`cad_share_drawing`	Uploads drawing to Cloudflare D1; returns real-time view & edit URLs.	`name`, `expiresDays: 7`
`cad_open_shared`	Fetches and loads a shared cloud drawing by ID or URL into the session.	`urlOrId: "https://autocad.site/?share=..."`
`cad_update_shared` / `cad_revoke_shared`	Updates or revokes shared drawings using editToken / ownerToken.	`id`, `token`

Shortcut	Command	Description

Turning an image or scan into a drawing

For users and AI agents (through the MCP server). A picture is only a picture: to edit it as CAD it has to be redrawn as lines, text, hatches and blocks.

Small sheets (forms, tables, details)

Attach the image (IMAGEATTACH, or drop the file on the drawing) as a reference.
Measure positions from the image pixels rather than by eye, so table lines and cells line up.
Draw on separate layers (frames, grid, fills, text, logo) next to or over the image; agents use cad_batch_add_entities.
Compare with the original (cad_export_svg for agents) and correct the differences.

Large plans (architectural drawings, many sheets of detail)

Reading a big plan in one look does not work: fine text and dimensions are lost. Split the work: algorithms find the geometry, the AI decides what it means, and the result is checked against the image.

Underlay and scale first. Attach the image on a locked layer. Use SCALE with Reference (or ALIGN) so a known dimension or the scale bar measures true. One wrong scale ruins everything after it, so confirm it.
Work in tiles. Cut the image into overlapping tiles (about 1000–1500 px), handle each one and place the results in drawing coordinates.
Geometry by code, not by eye. Image tools (for example OpenCV) clean the scan (straighten, black and white, remove specks), find straight lines and arcs (door swings), merge broken pieces, make near-horizontal and near-vertical lines straight, close gaps at corners, and pair parallel thick lines into walls (MLINE). The AI reviews and fixes the result.
Text by OCR. OCR (for example Tesseract or PaddleOCR, both read Arabic) reads room names and dimensions; the AI decides which text is a dimension, a room name or a note.
Blocks and layers. Doors, windows, sanitary fittings and furniture become one block each, inserted many times. Sort everything into layers such as walls, doors, windows, text and dimensions.
Build in batches. Agents send a few hundred objects per cad_batch_add_entities call and use cad_execute_script for procedural parts, so thousands of coordinates do not pass through the conversation. Save progress tile by tile (cad_export_dxf).
Check automatically. Export SVG, overlay it on the image, score each tile for how well the lines match, and redo only the tiles that score low.
Ask when unsure. Hand-drawn plans, blurry scans, hatching and stamps defeat automatic tracing. Keep the image underneath as a reference and ask instead of guessing.

What JsCAD does and does not do

JsCAD provides the drawing side: image underlay (embedded in the drawing), layers, blocks, multilines, tables, text, scaling by reference, DXF and SVG export, and the MCP tools above.
JsCAD does not trace images or read text from them automatically: steps 3 and 4 need outside image and OCR tools.
Expect automation to do most of a clean printed plan, with review for the rest; poor scans need much more manual work.

JsCAD — Browser-Based 2D CAD Platform

📐 Geometric Drafting Engine

⌨️ Command Line (AutoCAD-compatible)

🎯 Precision Snapping & Layers

📁 File Interchange & Multi-Space

JsCAD Generative CAD MCP Server & Browser Bridge

🚀 Why MCP & Generative CAD?

⚡ Key Superpowers for AI Agents

🏗️ System Architecture

🛠️ Complete MCP Tool Inventory (16 Tools)

🌐 Direct In-Browser AI Bridge

1. In-Page JavaScript API (window.CAD)

2. URL Query Parameters & Deep-Links

3. Live Zero-Refresh Hash Streaming

4. BroadcastChannel & PostMessage

⚙️ Setup with AI Agent Clients

1. Google Antigravity & Gemini CLI (~/.gemini/config/mcp_config.json)

2. Claude Desktop (claude_desktop_config.json)

3. Cursor IDE (.cursor/mcp.json)

⚡ Live Web Bridge Playground

Turning an image or scan into a drawing

Small sheets (forms, tables, details)

Large plans (architectural drawings, many sheets of detail)

What JsCAD does and does not do

1. In-Page JavaScript API (`window.CAD`)

1. Google Antigravity & Gemini CLI (`~/.gemini/config/mcp_config.json`)

2. Claude Desktop (`claude_desktop_config.json`)

3. Cursor IDE (`.cursor/mcp.json`)