Loading SaaS profile...
agentcad
Open-source CAD design and rendering tool for coding agents.
What it does
• Provides an MCP server and CLI allowing coding agents to execute, test, and debug 3D CAD scripts locally. • Automatically runs pre-execution validation checks to ensure generated geometries are watertight and dimensionally accurate. • Renders multi-angle PNG visual previews and topology reports for automated agent self-correction loops. • Exports versioned CAD models into standard manufacturing formats including STEP, STL, and GLB files.
Why it works
Local execution and MCP architecture integrate seamlessly into existing AI coding environments like Claude Code and Cursor without friction. Automated validation feedback loops allow coding agents to fix geometric errors before human review. Open-source distribution under the Apache-2.0 license removes cost barriers for individual makers and developers. Native support for build123d and CadQuery scripts provides robust parametric modeling primitives.
Category: Design & Creative
Team Size: 1
Visit Website
As of 2026
Design & Creative
Revenue: 0-$100k ARR
Open-source CAD design and rendering tool for coding agents.
Overview
agentcad is an open-source CAD design tool and MCP server built specifically for coding agents like Claude Code or Cursor. It bridges the gap between text-to-code generation and 3D modeling by allowing AI agents to write, execute, and validate build123d or CadQuery scripts locally. The platform automatically checks geometric integrity, renders visual previews, and outputs manufacturing-ready STEP, STL, and GLB files.
HQ: United States
Funding status: Bootstrapped
Revenue source: Open Source
Customer type: B2B
Pricing: Free (Open Source)
Tech stack: Python, MCP, build123d, CadQuery, CLI
Platform: Desktop • Web
Integrations: Claude Code, Cursor, Windsurf, Python, build123d, CadQuery
Founder story
agentcad was created by developer James Dillard as an open-source project to solve the inherent unreliability of AI-generated 3D models. Recognizing that coding agents frequently output broken geometry due to a lack of visual and programmatic feedback, he built an iterative command-line tool and MCP server that gives agents a tight validation loop.
What it does
- • Provides an MCP server and CLI allowing coding agents to execute, test, and debug 3D CAD scripts locally
- • Automatically runs pre-execution validation checks to ensure generated geometries are watertight and dimensionally accurate
- • Renders multi-angle PNG visual previews and topology reports for automated agent self-correction loops
- • Exports versioned CAD models into standard manufacturing formats including STEP, STL, and GLB files.
Who it's for
- Software Developers
- Mechanical Engineers
- 3D Printing Hobbyists
- AI Workflow Researchers
Why it works
- Local execution and MCP architecture integrate seamlessly into existing AI coding environments like Claude Code and Cursor without friction
- Automated validation feedback loops allow coding agents to fix geometric errors before human review
- Open-source distribution under the Apache-2.0 license removes cost barriers for individual makers and developers
- Native support for build123d and CadQuery scripts provides robust parametric modeling primitives.
Growth strategies
- Open- source distribution via GitHub with community-driven contributions and extensions
- Viral developer adoption through AI coding agent communities on X and Reddit
- Integration tutorials showcasing end- to-end multi-part assembly generation like the Wright Flyer
- Strategic alignment with modern AI developer tools and Model Context Protocol (MCP) registries
Alternatives
Comparison overview
- agentcad provides a programmatic, agent-driven CAD workflow using Python scripts and MCP servers [one Unlike traditional interactive GUI CAD packages like FreeCAD or Fusion 360, it is designed entirely for coding agents to self-validate and render 3D geometry programmatically n/]