The skill
Teach your AI to design well.
The PraxisCAD skill is a set of instructions, design rules and worked examples. Install it once in your AI client, and it knows how to model, edit, assemble, draw and check parts in PraxisCAD.
Install the skill
If your client has a plugin marketplace, add the skill with the two commands below. For other clients, upload praxiscad-skill.zip from the latest release of the skill repository.
/plugin marketplace add praxiscad/praxiscad-skill
/plugin install praxiscad@praxiscadPraxisCAD also sends core instructions through its MCP connection, so the skill is optional. It improves design rules, hole and thread workflows, assemblies, drawings and manufacturability checks. Update it when you update PraxisCAD.
Where it goes
Any MCP-capable AI client. Pick the way your client takes instructions.
Skill folders
Upload praxiscad-skill.zip from the latest release, or copy skills/praxiscad into your client's skills directory.
Project instructions
Add instructions/AGENTS.md to your client's project instructions or rules.
System prompts
Paste instructions/AGENTS.md into your client's system prompt.
The skill is not connected to your AI by itself. Connect your client to PraxisCAD first (Connect your AI), then add the skill.
No PraxisCAD installed yet?
The skill tells your AI what to do when PraxisCAD is missing: it checks for the app, shows you the download link for your computer, and asks before it downloads or installs anything. You can also use the links yourself. They always point to the latest release.
- macOS (Apple silicon and Intel): /get/macos
- Windows installer: /get/windows
- Windows MSI (IT deployment): /get/windows-msi
Or use the download page to see the current version and release notes.
What is inside
Modelling workflows from text or photos, a hole wizard and threads, sketch tools and DXF, materials and renders, editing of imported STEP files, assemblies and motion, 2D drawings with tolerances and GD&T, design checks for CNC, 3D printing, injection moulding and sheet metal, plus worked examples. It is model-agnostic and free; you use your own AI subscription.