web-search-mcp
web-search-mcp
为本地大语言模型(LLM)提供外部网页搜索和页面抓取功能,以 MCP 工具和 CLI 的形式呈现。该实现优先使用 Playwright,采用异步方式,默认启动系统安装的 Chromium 二进制文件。设计历史记录在 docs/IMPLEMENTATION_PLAN.md 中。
工具
search(provider, context)在duckduckgo(默认)、google或yandex上运行基于浏览器的搜索,并返回最多 5 个自然搜索结果的 URL。fetch(urls)抓取1..5个 URL,并返回浏览器渲染后的 HTML,包含每个 URL 的pages、errors和truncated字段。
context 特意设计为搜索查询字符串,以保持调用者兼容性。
Related MCP server: Web Search MCP
设置
uv sync
chromium --version如果 Chromium 安装在非标准位置,请设置 PLAYWRIGHT_CHROMIUM_EXECUTABLE=/path/to/chromium。
使用方法
通过 stdio 运行 MCP 服务器:
uv run web-search serve-mcp直接运行 CLI:
uv run web-search search --context "python async playwright"
uv run web-search fetch https://example.com暴露 HTTP 传输而非 stdio:
uv run web-search serve-mcp --transport streamable-http --host 127.0.0.1 --port 8000设计说明
每个进程共享一个浏览器,每个请求使用一个新的无痕上下文。
Playwright 启动系统 Chromium 二进制文件,而不是 Playwright 下载的浏览器包。
系统 Chromium 可与 Playwright 配合使用,而测试的系统 Firefox 构建在启动后立即退出。因此,本仓库将系统 Chromium 作为支持的主机浏览器路径。
浏览器冒烟测试应从普通主机 shell 运行。受限沙箱即使在主机浏览器工作正常时也可能阻止 Chromium 启动。
全局导航并发上限为
3。超时设置:每页
15s,search总计20s,fetch总计35s。SSRF 防护:仅限
http/https,无嵌入式凭据,在导航前和浏览器子请求时阻止回环/私有/链路本地/保留 IP。JavaScript 挑战页面有
10s的有限稳定窗口;不包含验证码破解、隐身指纹识别或特定站点的绕过逻辑。HTML 每个 URL 上限为
1 MiB;超大的响应会被截断并在truncated中报告。v1 版本不参考
robots.txt。
开发
source .venv/bin/activate
pytest解析器测试针对保存的 HTML 固定数据运行;选择器漂移是预期的维护成本。
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Web search, browser automation, scraping, crawling and CAPTCHA solving for AI agents.
A real browser for your agent: screenshots, page checks, Markdown and PDFs.
Hosted browser for agents: open pages, search the web, extract data, read PDFs and take screenshots.
Web data for agents. Extract web pages, crawl websites, discover URLs, and search the web.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables LLMs to interact with web pages, take screenshots, and execute JavaScript in a real browser environment10297MIT
- AlicenseNot gradedqualityDmaintenanceEnables web searching through Google, DuckDuckGo, and Bing using a headless Chrome browser, returning structured results with titles, URLs, and snippets. Also supports fetching and extracting text content from any webpage.16MIT
- AlicenseAqualityDmaintenanceProvides web access capabilities for LLMs including search, fetching, content extraction, PDF reading, image viewing, and screenshots.333 PyPI46MIT
- FlicenseNot gradedqualityDmaintenanceEnables AI agents to perform web searches, extract webpage content, and conduct end-to-end search-and-extract operations using multiple search providers and content extraction methods.-