在 Claude Code 里搭建MCP服务器：一帖专治走投无路的技术“玄学”

bash <<'EOF'
echo "🔧  正在安装 Claude MCP 服务器（最新版）…"

# 顺序思维 — Claude 的“思考链”引擎
claude mcp add sequential-thinking -s user \
  -- npx -y @modelcontextprotocol/server-sequential-thinking || true

# 文件系统 — 赋予 Claude 访问本地文件夹的权限
claude mcp add filesystem -s user \
  -- npx -y @modelcontextprotocol/server-filesystem \
     ~/Documents ~/Desktop ~/Downloads ~/Projects || true

# Playwright — 现代化的跨浏览器自动化工具
claude mcp add playwright -s user \
  -- npx -y @playwright/mcp-server || true

# Puppeteer — 仅支持 Chrome（已弃用，但还能用）
claude mcp add puppeteer -s user \
  -- npx -y @modelcontextprotocol/server-puppeteer || true

# Fetch — 简单的 HTTP GET/POST
claude mcp add fetch -s user \
  -- npx -y @kazuph/mcp-fetch || true

# 浏览器工具 — 开发者工具的日志、截图等
claude mcp add browser-tools -s user \
  -- npx -y @agentdeskai/browser-tools-mcp || true

echo "--------------------------------------------------"
echo "✅  MCP 注册完成。"
echo ""
echo "🔴  要启用“浏览器工具”，请在*另一个*终端里运行以下命令，并保持该窗口开启："
echo "    npx -y @agentdeskai/browser-tools-server"
echo "--------------------------------------------------"
claude mcp list
EOF

claude mcp add sequential-thinking -s user \
  -- npx -y @modelcontextprotocol/server-sequential-thinking

claude mcp add filesystem -s user \
  -- npx -y @modelcontextprotocol/server-filesystem \
     ~/Documents ~/Desktop ~/Downloads ~/Projects

claude mcp add playwright -s user \
  -- npx -y @playwright/mcp

claude mcp add puppeteer -s user \
  -- npx -y @modelcontextprotocol/server-puppeteer

claude mcp add fetch -s user \
  -- npx -y @kazuph/mcp-fetch

npx -y @agentdeskai/browser-tools-server@1.2.1

claude mcp add browser-tools -s user \
  -- npx -y @agentdeskai/browser-tools-mcp@1.2.1

# 把 YOUR_API_KEY_HERE 换成你自己的 Brave Search API 密钥
claude mcp add brave-search -s user \
  -- env BRAVE_API_KEY=YOUR_API_KEY_HERE \
     npx -y @modelcontextprotocol/server-brave-search

# 把 fc-YOUR_API_KEY 换成你自己的 Firecrawl API 密钥
claude mcp add firecrawl -s user \
  -- env FIRECRAWL_API_KEY=fc-YOUR_API_KEY \
     npx -y firecrawl-mcp

FIRSTLY, READ THROUGH THIS PROMPT THOROUGHLY. UNDERSTAND ALL ASPECTS. Using sequentialthinking, find ONE segment ranked with the highest priority to take action upon within the optimized-segments.md and cross reference the chosen segments to work on from the Segment's directive against its correlate documentation within Master Checklist and the Master Codebase files to ensure your directive within the optimized-segments.md is sound and valid. Then use sequentialthinking to push the necessary, comprehensive changes as directed to the codebase. Before, during, and after codefix and file changes you will follow the directives of the optimized-segments.md respecting its guidelines and formatting and precision -- and then UPDATE THE STATUSES to the appropriate SECTION(s) of the optimized-segments.md Master Checklist.md and the Master Codebase.md (keep concise and bloat free, and ensure your updates to these highly important documents are only to relevant sections using edit_file functionality and not complete re-writes).

“首先，把整个提示词给我通读一遍，每个细节都要吃透。运用顺序思维，在 optimized-segments.md 文件里，找到优先级最高的一个任务模块来执行。然后，将这个选定模块的指令与 Master Checklist.md 和 Master Codebase.md 文件里的相关文档进行交叉比对，确保 optimized-segments.md 里的指令是合理且有效的。接着，再次运用顺序思维，按照指令对代码库进行必要且全面的修改。在修改代码和文件的整个过程中——无论是之前、之中还是之后——你都必须遵循 optimized-segments.md 里的指令，严格遵守它的规范、格式和精度要求——然后，更新 optimized-segments.md、Master Checklist.md 和 Master Codebase.md 这几个文件里相应板块的状态。（注意保持简洁，别写废话，并且确保你只用编辑功能修改这些重要文件的相关部分，绝对不要整个文件重新写一遍）。“