OpenCoworkAI/open-codesign

Open-source Claude Design alternative. One-click import your Claude Code / Codex API key. Prompt → prototype / slides / PDF. Multi-model (Claude, GPT, Gemini, Kimi, GLM, Ollama). BYOK, local-first, MIT.

What it solves

Commercial AI design tools such as Claude Design, v0 by Vercel, Lovable, and Bolt.new tend to lock you into one model provider, one cloud workspace, and a recurring subscription. Open CoDesign attacks all three at once: it is an open-source, MIT-licensed desktop app that runs on your own laptop, works with whatever model you already have a key or subscription for, and keeps your designs and session history on-device. The result is AI-assisted design without mandatory cloud processing, provider lock-in, or extra subscription fees beyond what you already pay for model access.

How it works

You type a prompt (or pick one of fifteen built-in demos such as landing page, dashboard, pitch slide, mobile app, or chat UI) and an agent plans, writes, self-checks, and produces a sandboxed interactive prototype — with hover states, tabs, and empty states already wired up. Everything runs in a desktop app built on Electron. The agent's work stays visible: a live panel shows todos and streaming tool calls, and you can interrupt generation at any point.

You connect a model in one of three ways: sign in with a ChatGPT subscription to use Codex models, paste an API key for Anthropic, OpenAI, Gemini, OpenRouter, DeepSeek, or similar providers, or use a local keyless setup like Ollama. Existing Claude Code or Codex provider configurations import in one click. Credentials stay in local config files and nothing leaves your machine unless your chosen model route requires it.

Beyond generation, the app offers targeted refinement: you can click any element, drop a pin, and ask the model to rewrite just that region; or use AI-emitted sliders to fine-tune color, spacing, and typography without writing a new prompt. Every iteration is saved locally with instant switching between recent versions. Finished work exports as real, editable files — HTML, PDF, PPTX, ZIP, or Markdown.

Who it's for

Developers, designers, and product people who want the speed of AI-native design tools but refuse cloud-only workflows, single-provider lock-in, or subscription pricing. It also suits anyone who already pays for a model (API key or ChatGPT subscription) and wants to put that spend to work in a local, private environment — and teams that want version history and editable exports rather than opaque cloud artifacts.

Highlights

  • Runs entirely on your laptop — no mandatory cloud workspace and no cloud processing of your designs
  • Works with 20+ models including Claude, GPT, Gemini, DeepSeek, Ollama, and any OpenAI-compatible endpoint
  • Three connection paths: ChatGPT subscription sign-in, paste-in API key, or local keyless Ollama/gateway
  • One-click import of existing Claude Code or Codex provider configs — running in under 90 seconds
  • Comment mode: click any element, drop a pin, and let the model rewrite only that region
  • AI-tuned sliders that surface the parameters worth tweaking for color, spacing, and typography
  • Local version history with instant switching between recent design iterations
  • Live, interruptible agent panel showing todos, tool calls, and progress
  • Exports editable files: HTML, PDF, PPTX, ZIP, and Markdown
  • Fifteen built-in demo prompts — landing page, dashboard, pricing, chat UI, and more
  • MIT-licensed open source with installers for macOS, Windows, and Linux

Related

  • Project
    nexu-io/open-design

    OpenDesign is a desktop‑first, Apache‑2.0‑licensed workspace that lets teams use any installed LLM coding agent (Claude, Codex, DeepSeek, etc.) to generate and iterate on design artefacts—web prototypes, slide decks, dashboards, images, and motion‑graphics videos. It ships with hundreds of design‑system packages and template libraries, supports 26+ agent CLIs via a simple plug‑in system, runs locally (macOS/Windows, source‑build for Linux), and can export results as HTML, PDF, PPTX, or MP4 via the open‑source HyperFrames engine.

  • Project
    OpenCoworkAI/open-cowork

    Open Cowork is a free, open‑source desktop AI‑agent app for Windows and macOS. It wraps Claude‑style models (Claude, OpenAI, Gemini, Chinese models) in a GUI, offers VM‑level sandboxing, a built‑in Skills system for Office‑document generation, MCP connectors for browsers/Notion, GUI automation, and remote control via Feishu or Slack. One‑click installers and Homebrew casks make setup trivial; developers can also build from source.

  • Project
    gszhangwei/open-spdd

    A CLI tool and methodology that transforms AI coding prompts into structured design contracts using the REASONS Canvas framework to ensure synchronization between design and implementation.

  • Project
    anomalyco/opencode

    OpenCode is an open-source AI coding agent available as a CLI and a beta desktop app. It installs via curl script, npm, Homebrew, pacman, mise, or Nix. It includes two built-in agents — "build" (full-access, default) and "plan" (read-only) — switchable with Tab, plus a "general" subagent. Docs are at opencode.ai/v2/docs, and projects using "opencode" in their name must disclaim affiliation with the team.

  • Project
    NanmiCoder/open-image-prompts

    An open-source, local-first visual prompt archive and agent-ready toolkit for retrieving traceable image-prompt references and defining art direction.