Esta página se muestra en inglés. Una traducción al español está en curso.
ConfigsJun 2, 2026·3 min de lectura

Open CoDesign — Open-Source AI Design Tool with Multi-Model Support

A local-first, open-source alternative to commercial AI design tools that generates prototypes, slides, and PDFs from prompts using Claude, GPT, Gemini, or local models via Ollama.

Listo para agents

Instalación con revisión previa

Este activo requiere revisión. El prompt copiado pide dry-run, muestra escrituras y continúa solo tras confirmación.

Needs Confirmation · 66/100Política: confirmar
Superficie agent
Cualquier agent MCP/CLI
Tipo
Skill
Instalación
Single
Confianza
Confianza: Established
Entrada
Open CoDesign Overview
Comando con revisión previa
npx -y tokrepo@latest install b894a9e1-5e7d-11f1-9bc6-00163e2b0d79 --target codex

Primero dry-run, confirma las escrituras y luego ejecuta este comando.

Introduction

Open CoDesign puts AI-powered design generation on your own machine. Instead of sending designs through third-party services, you bring your own API key for any supported model and generate UI prototypes, presentation slides, and exportable documents locally. The MIT-licensed codebase ensures full transparency and customization.

What Open CoDesign Does

  • Generates web, desktop, and mobile UI prototypes from text prompts
  • Creates presentation slides with consistent visual themes
  • Exports to HTML, PDF, PPTX, and MP4 formats
  • Supports multiple LLM providers including Claude, GPT, Gemini, Kimi, GLM, and Ollama
  • Provides a desktop app via Electron and a browser-based interface

Architecture Overview

Open CoDesign is built as a TypeScript Electron app with a Vite-powered web frontend. The prompt engine parses design requests and routes them to the selected LLM provider. Generated designs render in a sandboxed preview iframe. A template system provides reusable design components and layout primitives. Export pipelines convert the rendered output to various formats using Puppeteer for PDF and HTML capture.

Self-Hosting & Configuration

  • Requires Node.js 18+ for the web version or the Electron binary for desktop
  • Bring your own API key for any supported provider via the settings panel
  • For offline use, configure a local Ollama endpoint with a capable model
  • Design templates are stored as JSON files and can be customized or extended
  • All generated designs stay on your machine unless you choose to export

Key Features

  • BYOK (Bring Your Own Key) for any supported LLM provider
  • Local-first architecture with no data leaving your machine
  • Multi-format export (HTML, PDF, PPTX, MP4)
  • Cross-platform desktop app and web interface
  • Extensible template system for consistent design output

Comparison with Similar Tools

  • Claude Design — proprietary and cloud-only; Open CoDesign is open source and local-first
  • Figma — manual design tool; Open CoDesign generates designs from prompts
  • v0 by Vercel — generates React components; Open CoDesign targets broader design artifacts including slides and PDFs
  • Penpot — collaborative design tool; Open CoDesign focuses on AI generation rather than manual drawing

FAQ

Q: Which LLM works best for design generation? A: Claude and GPT-4 class models produce the highest quality designs. Smaller local models work for simpler layouts but may struggle with complex multi-component pages.

Q: Can I use my company design system? A: Yes, you can add custom CSS themes and component templates that the AI will use as building blocks.

Q: Is the generated code production-ready? A: Generated HTML and CSS is clean and semantic, but treating it as a starting point for production rather than final output is recommended.

Q: Does it work fully offline? A: Yes, with Ollama or another local LLM provider, the entire workflow runs without internet access.

Sources

Discusión

Inicia sesión para unirte a la discusión.
Aún no hay comentarios. Sé el primero en compartir tus ideas.

Activos relacionados