mobbin-forge
An MCP server that works with Mobbin — 600k+ real app screens and flows from 1,100+ apps. Search screens, browse flows, grab palettes, build better UI without leaving Claude.
README
mobbin-forge
The original mobbin-mcp is archived and broken. This fork fixes it.
An MCP server that works with Mobbin — 600k+ real app screens and flows from 1,100+ apps. Search screens, browse flows, grab palettes, build better UI without leaving Claude. Mobbin has no public API. This server reverse-engineers their internal endpoints.
This works. Fork tested live 2026-07-14. Unwraps Mobbin's new .value wrapper, handles type drift (popularApps now category-grouped), confirmed 893 apps + 42 categories running. Ground truth API responses in repo.
What works right now
✅ getSearchableApps(platform) — 893 real apps (Disney+, ChatGPT, Arc Search...)
✅ getPopularApps(platform) — 42 categories grouped (ai, finance, crypto, productivity...)
✅ getAppScreens(appId, platform) — the target app's full structured screen set, including patterns and UI elements. Pixel-verified on ChatGPT (384 screens), Claude (107), and Perplexity (131). Point at an app, get the right app — not recommendation noise.
✅ Image downloads — CDN URLs via Bytescale (public, no auth, tested with real webp files)
✅ Auth — Session cookie stored chmod 600 in ~/.mobbin-mcp/auth.json, free account works
11 MCP tools: search apps/screens/flows, quick search, get app screens/flows, popular apps, collections, screen detail + color extraction, filters.
One limitation
Global cross-app search moved to client-side rendering (Next.js SPA state), so it's not directly callable yet. Everything else — per-app screens, categories, downloads — works great. You browse app-by-app, which is how you'd design anyway. Restoration playbook in investigate-search-routes.md if you want to hack on it.
Setup
Prerequisites
- Node.js 18+
- A Mobbin account (free works)
1. Install & build
git clone https://github.com/Jelil-ah/mobbin-forge.git
cd mobbin-forge
npm install
npm run build
2. Authenticate
npx mobbin-mcp auth
Follow prompts: open mobbin.com, log in, run copy(document.cookie) in browser console, paste into CLI. Session saved to ~/.mobbin-mcp/auth.json (chmod 600), auto-refreshed. For env-var auth, see CONTRIBUTING.md.
3. Add to Claude Code
claude mcp add mobbin -- npx -y mobbin-mcp
Claude Desktop / Codex: add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"mobbin": {
"command": "npx",
"args": ["-y", "mobbin-mcp"]
}
}
}
Example prompts
- "Show me how DoorDash and Uber Eats handle checkout flows"
- "Pull up Duolingo's onboarding and walk me through the design decisions"
- "Find login screens using bottom sheets and extract color palettes"
How it works
Calls Mobbin's internal API routes using your session cookie. Tokens auto-refresh via Supabase and persist to ~/.mobbin-mcp/auth.json. getAppScreens resolves the target app's real version, then parses its structured Next.js RSC data chunk — screens, patterns, elements, and dimensions — before downloading images through Bytescale. Optional color extraction uses sharp.
Credits
Forked from pdcolandrea/mobbin-mcp. Respect to the original — pioneered reverse-engineering Mobbin's API. This fork exists because upstream is archived and the API changed.
License
ISC
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。