TechDraw AI Skills

AI agent skills for turning a photograph of a physical part into engineering documentation with TechDraw AI. Draft a dimensioned 2D technical drawing, trace it into CAD-ready DXF and DWG, reconstruct the part as a 3D mesh, or render it as a finished product shot. Works with Claude Code, Claude Desktop, Cursor, Codex, VS Code, Windsurf and anything else that loads Markdown skills or speaks MCP.

Each user authenticates with their own TechDraw AI API key, and every generation is billed to their account credits. The skills call the public TechDraw AI API directly, so there is nothing extra to install and no CAD kernel to build.

_{One photograph, three outputs: a dimensioned drawing, CAD geometry, and a 3D model.}

Install

Pick one. Each method sets the skills up for your agent.

npx skills (recommended, cross-agent)

npx skills add MeltFlexDevs/techdrawai-skills

GitHub CLI (v2.90+)

gh skill install MeltFlexDevs/techdrawai-skills

Claude Code marketplace

Inside Claude Code:

/plugin marketplace add MeltFlexDevs/techdrawai-skills
/plugin install techdrawai@techdrawai

MCP server only

If you want the tools without the skills, add the MCP server directly:

claude mcp add techdrawai -- npx -y techdrawai-mcp

Configuration for Cursor, Claude Desktop, Codex, VS Code and Windsurf is on the MCP page.

Setup script

Universal fallback:

git clone --depth 1 https://github.com/MeltFlexDevs/techdrawai-skills.git
cd techdrawai-skills
./setup

More options in INSTALL.md. Agent-driven install, to paste into your agent: INSTALL_FOR_AGENTS.md.

Authenticate

Programmatic access is part of the Studio plan. Once you are subscribed, the simplest way to sign in is the browser flow, which mints a key from your session and hands it back over loopback so you never copy or paste a secret:

npx -y techdrawai-mcp auth login

Or create one by hand at techdrawai.com/app → Settings → API keys and export it:

export TECHDRAWAI_API_KEY="tda_sk_..."

The key is shown exactly once at creation. Only its SHA-256 is stored, so nobody can recover it afterwards, including us. An account may hold five live keys and revoking one takes effect on the next request.

Skills

Tools

Failed generations are refunded automatically, so an agent that retries a bad prompt does not quietly drain the balance.

Every tool writes its result to disk (./techdrawai-output/ unless you pass output_path) and returns the path rather than the bytes. A technical drawing is a megabyte-scale image, and pushing that through the protocol would burn the agent's context on something it usually just hands to the next tool.

What it looks like in practice

You: Here is a bracket, ~/photos/bracket.jpg. It is 120 mm across. I need a DXF for the laser cutter.

[identify_product]  → "steel wall bracket"

[generate_technical_drawing]
  product_name: "steel wall bracket"
  measurements: [{ value: "120", unit: "mm", position: "overall width" }]
  standard: ISO · paper_size: A3 · drawing_type: working
→ ./techdrawai-output/drawing-2026-07-27_15-04-11.png   3 credits

[image_to_dxf]  width_mm: 120
→ ./techdrawai-output/drawing-2026-07-27_15-04-11.dxf   34 contours, free

The DXF arrives at true size, so it goes straight to the cutter without rescaling. More patterns in COOKBOOK.md.

One measurement changes everything

Without a real measurement the dimensions are the model's best estimate and are marked nominal. With even one, every other feature is scaled proportionally from it and the whole drawing becomes internally consistent and manufacturable. If a user has not given one, ask for it. It is the single biggest quality lever in the whole workflow.

The CLI

The same package is a command-line tool when you run it with arguments instead of letting an agent speak MCP to it:

techdrawai draw -i bracket.jpg --measure 120:"overall width" --paper A3
techdrawai dxf  -i drawing.png --width-mm 120 --dwg
techdrawai model3d -i bracket.jpg --detail high

Full reference at techdrawai.com/cli.

Documentation

• API reference: every endpoint, request and response
• CLI: every command and flag
• MCP: client-by-client setup
• COOKBOOK.md: worked examples
• CONTRIBUTING.md

Licence

MIT. See LICENSE.