Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者提供的说明(英文原文)
195 UX/UI principles with citations, 12 UX smells and 12 flow checklists for AI agents, EN and ES.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
UXUI_API_KEY标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"uxui-principles": {
"command": "npx",
"args": [
"-y",
"@uxuiprinciples/mcp-server"
],
"env": {
"UXUI_API_KEY": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add uxui-principles --env UXUI_API_KEY=your-value -- npx -y @uxuiprinciples/mcp-server运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"uxui-principles": {
"command": "npx",
"args": [
"-y",
"@uxuiprinciples/mcp-server"
],
"env": {
"UXUI_API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"uxui-principles": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@uxuiprinciples/mcp-server"
],
"env": {
"UXUI_API_KEY": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月8日
发布者(joselhurtado)不是仓库所有者(uxuiprinciples);可能是其他人代为上架的
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 0 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:UXUI_API_KEY
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(1.2.1)于昨天进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Node.js。
同一分类中有:Spec Driven Development, Remote Desktop Commander, Readfirst
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:↓ 每周下载量i★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
What to read and check before changing a file: project rules, risks, checks and tests per file.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
MCP server for UX/UI Principles: 195 research-backed principles with their citations, 12 UX smells with fix recipes and 12 user-flow checklists, in English and Spanish, inside Claude Code, Claude Desktop, Cursor, VS Code, Windsurf or any MCP client.
Free and offline by default: the dataset ships inside the package and the server makes no network call. A Principles Library API key turns on pro mode, which adds the full guide bodies, the AI prompts and an interface audit through the uxuiprinciples.com API.
Model Context Protocol (MCP) is an open protocol that lets an AI assistant call external tools. This server speaks MCP over stdio and registers 13 tools.
Requires Node.js 18 or newer. npx -y fetches the latest version on first run.
claude mcp add uxui-principles -- npx -y @uxuiprinciples/mcp-server
Add --scope user to make it available in every project. Confirm with claude mcp list.
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop:
{
"mcpServers": { "uxui-principles": { "command": "npx", "args": ["-y", "@uxuiprinciples/mcp-server"] } } }
Add to .cursor/mcp.json in the project (or ~/.cursor/mcp.json for every project), then enable the server in Cursor Settings, MCP:
{
"mcpServers": { "uxui-principles": { "command": "npx", "args": ["-y", "@uxuiprinciples/mcp-server"] } } }
Add to .vscode/mcp.json in the workspace, then start the server from the MCP panel or let Copilot agent mode start it:
{
"servers": { "uxui-principles": { "type": "stdio", "command": "npx", "args": ["-y", "@uxuiprinciples/mcp-server"] } } }
Add to ~/.codeium/windsurf/mcp_config.json, then refresh the MCP list in Cascade:
{
"mcpServers": { "uxui-principles": { "command": "npx", "args": ["-y", "@uxuiprinciples/mcp-server"] } } }
npm install -g @uxuiprinciples/mcp-server
uxui-mcp
Point the client at the uxui-mcp binary instead of npx. Re-run the install to pick up new versions.
The Principles Library ($59/year) includes a REST API key. Give it to the server as the UXUI_API_KEY environment variable and four tools change:
| Tool | Free mode | Pro mode |
|---|---|---|
get_principle | Metadata, summary, citations, link | The same plus the full guide body in content |
get_principle_prompts | Link to the guide | The principle's AI prompts |
audit_interface | Refuses with the key hint | Findings mapped to principles and smells, score, priority fixes |
check_requirements | Refuses with the key hint | Pass or fail per required-state rule, each with its principle and citations |
Everything else keeps running offline. If the API is unreachable, get_principle and get_principle_prompts fall back to the free answer and report the failure in proError; audit_interface and check_requirements return the error code.
{
"mcpServers": { "uxui-principles": { "command": "npx", "args": ["-y", "@uxuiprinciples/mcp-server"], "env": { "UXUI_API_KEY": "uxui_live_..." } } } }
For Claude Code: claude mcp add uxui-principles -e UXUI_API_KEY=uxui_live_... -- npx -y @uxuiprinciples/mcp-server. The key is read from the environment only and is never logged. Get or rotate yours at uxuiprinciples.com/account. The server's startup line on stderr says pro mode or free mode, and the instructions it hands the client say the same.
| Tool | Description |
|---|---|
get_principle | One principle by slug: summary, citations (author, title, year, url) and the link to the full guide. Pro: the full guide body |
search_principles | Search by query, part, difficulty, or tags |
list_principles | List all principles or filter by part |
get_principle_prompts | The AI prompts for a principle. Pro: the prompt texts; free: the link to the guide |
| Tool | Description |
|---|---|
get_smell | Get a smell with symptoms and fix recipe |
list_smells | List all smells, optionally by category |
detect_smell | Detect smells from a UI description, written in English or Spanish |
suggest_fix | Get fix suggestions for a UX issue |
| Tool | Description |
|---|---|
get_flow | Get a flow with pre/post-flight checklists |
list_flows | List all available flows |
generate_checklist | Generate checklist for a flow phase |
get_flow_principles | Get key principles for a flow |
| Tool | Description |
|---|---|
audit_interface | Audit a described screen (20 to 5,000 characters): findings with severity, remediation and business impact, score, interface type, strengths and the three priority fixes. Needs UXUI_API_KEY |
check_requirements | Check a screen given as data (flow, components, actions with consequential and automated flags, states implemented) against eight required-state rules: confirmation before consequential actions, error, empty and loading states, undo or edit, sources on AI output, recovery after an AI error, consent before automation. Pass, fail or not_applicable per rule with the principle and its citations, plus a markdown summary. Needs UXUI_API_KEY |
You: What principle should I follow for form design?
Claude: Let me search for form-related principles... [Calls search_principles with query="form"]
Based on the UX/UI Principles database, here are the key principles for form design:
...
You: My login form has a 60% abandonment rate. Help me fix it.
Claude: Let me detect potential UX issues... [Calls detect_smell with description="login form high abandonment"]
I detected "Form Graveyard" smell with high confidence. Let me get the fix recipe... [Calls get_smell with id="form-graveyard"]
Here's my analysis and recommendations:
...
Let me also get the flow checklist for the form... [Calls generate_checklist with flow_id="forms" phase="postflight"]
With a key:
You: Audit this checkout: three steps, no order summary until the last one, the Pay button is disabled with no explanation until every field validates.
Claude: [Calls audit_interface with the description] Score 58, fair. Two warnings: Visibility of System Status (order summary hidden) and Silent Errors (disabled button without feedback). Priority fix: ...
After building a screen:
You: Build the settings page: profile form, an AI summary of recent activity, and a danger zone with "Delete account". Then check it.
Claude: [Builds the page, then calls check_requirements with flow "settings", the components, the actions (Delete account flagged consequential) and the states it implemented] 3 of 6 applicable rules fail: undo-or-edit (User Control and Freedom, Nielsen 1994), show-sources (AI Source Citations, Liu et al. 2023), ai-error-recovery (Efficient AI Correction, Amershi et al. 2019). Adding an undo toast, a sources line under the summary and a regenerate control...
All tools accept locale: "en" | "es". detect_smell and suggest_fix understand descriptions written in either language whatever the locale:
search_principles({ query: "navegación", locale: "es" })
detect_smell({ description: "el formulario es demasiado largo", locale: "es" })
get_principle (the guides cite 2,400+ academic citations in full)git clone https://github.com/joselhurtado/uxuiprinciples-web.git
cd uxuiprinciples-web npm install npm run build --workspace=@uxuiprinciples/principles-data npm run build --workspace=@uxuiprinciples/mcp-server node packages/mcp-server/dist/index.js
npm run smoke-test --workspace=@uxuiprinciples/mcp-server packs the build, installs it in a temp project and drives both modes over stdio (pro mode against a local fake of the API). server.json next to this README is the MCP Registry manifest; publish with mcp-publisher publish after the npm release.
MIT © UX/UI Principles