Qcad Mcp

QCAD MCP

[](https://github.com/sandraschi/qcad-mcp/actions) AI-driven 2D CAD automation — DXF/DWG parsing, 3D extrusion, ECMAScript scripting, native rendering, dimensions, annotations, and agentic workflows. 22 MCP tools. Your AI assistant becomes a QCAD Pro operator.

| | | |--:|--| | You might use this if… | You want your AI to write and execute QCAD scripts, process floor plans programmatically, convert DXF to 3D for Resonite/Unity3D, or automate CAD drafting without opening QCAD. | | What it connects to | ezdxf (free Python engine) for parsing/extrusion + QCAD Pro (~€42) for ECMAScript bridge, native rendering, DWG | | Ports | Backend 10966, Dashboard 10967 | | Start | just bootstrap then start.ps1 |

Two-Tier Engine

| Tier | Cost | Powers | |------|------|--------| | ezdxf | Free | DXF parsing, SVG preview, STL extrusion, room analysis, DXF creation from primitives, entity/layer modification, block search | | QCAD Pro | ~€42 | ECMAScript bridge (AI writes QCAD scripts), native SVG/PDF/BMP (perfect hatches/fonts/dims), dimensions, measurements, text annotations, hatch patterns, agentic workflows, DWG import/export |

QCAD Pro Features

With QCAD Pro 3.x installed the AI agent gains full CAD automation:

• plan_script — Execute arbitrary ECMAScript against DXF documents. Full QCAD API: entities, layers, blocks, operations. The AI writes code, QCAD Pro runs it headless. No script library needed — the LLM is the library.
• plan_agentic — Natural language CAD goals → ECMAScript → executed. "Create a 10m×8m floor plan with 4 rooms and dimensions" → done. Unlike hunting for a 20-year-old AutoLISP routine on a forum.
• plan_render — Native SVG/PDF/BMP output with correct hatches, TrueType fonts, lineweights, dimension styles.
• plan_dimension — Add aligned, radial, diametric, angular dimensions to drawings.
• plan_measure — Per-entity distance, angle, area, perimeter measurement via QCAD's geometry engine.
• plan_text / plan_hatch — Text annotations with full styling + hatch/fill patterns (ANSI31-38, AR-CONC, SOLID).
• plan_convert — DWG↔DXF conversion via Teigha engine.

Documentation Index

| Guide | Content | | :--- | :--- | | [Installation](docs/install.md) | Prerequisites, ezdxf, QCAD Pro setup, bootstrap | | [Architecture](ARCHITECTURE.md) | Two-tier engine, 20-tool table, service layer, pipeline | | [FreeCAD Pipeline](docs/freecad-pipeline.md) | NL→DXF→STL→BIM→IFC cross-repo chain with freecad-mcp | | [MCP Tools](docs/mcp-tools.md) | All 22 tools with examples, return formats | | [AI Tooling](docs/ai-tooling.md) | Ollama CAD chat, agentic plan analysis workflows | | [About QCAD](docs/about-qcad.md) | History, scripting API, ezdxf vs QCAD Pro, AutoCAD comparison | | [QCAD Pro vs. AutoCAD LT](docs/qcad-pro-vs-autocad-lt.md) | Feature comparison, scripting, pricing, platform support | | [Webapp README](webapp/README.md) | 12-page React dashboard: depot, viewer, extrude, blocks, layers |

Quick Start

# 1. Bootstrap
just bootstrap   # uv sync + npm install

# 2. (Optional) Set QCAD Pro path if installed
$env:QCAD_PRO_PATH = "C:\Program Files\QCAD"

# 3. Launch
start.ps1        # kills zombies, starts backend + frontend, opens browser

MCP Client Config

{
  "mcpServers": {
    "qcad": {
      "url": "http://localhost:10966/sse",
      "transport": "sse"
    }
  }
}

Once connected, call qcad_status to verify QCAD Pro availability, then plan_info on uploaded files, or plan_script to execute CAD operations.

Industrial Quality Stack

• Python (Core): Ruff for linting. 22 MCP tools on FastMCP 3.2. Zero lint errors.
• Webapp (UI): Biome + tsc for formatting and type safety. Zero errors across 18 files.
• Protocol: FastMCP 3.2 SSE transport + REST API (20+ endpoints).
• Automation: [Justfile](./justfile) recipes for all fleet operations.

License

MIT — see [LICENSE](LICENSE).

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: sandraschi
• Source: sandraschi/qcad-mcp
• License: MIT

Install and usage instructions live in the source repository linked above.