QuanticData MCP server
QuanticData MCP 服务器 — 让 Claude、Cursor 及任何 MCP 客户端获得实时网页访问能力
一个单文件、零依赖的 MCP 服务器,将 QuanticData 数据 API 置于 AI 代理之前:抓取页面、运行 Google/Bing/DuckDuckGo/Yandex 搜索、绘制或爬取网站地图、对 URL 进行 SEO 审计,并使用语义输入运行 31 个现成的 Collectors 中的任意一个。
无需 SDK、无需构建步骤、无需 node_modules。只需 Node.js 18+ 和一个文件。
git clone https://github.com/quantumproxies/quanticdata-mcp-server
cd quanticdata-mcp-server
QUANTICDATA_API_KEY=qd_live_your_key_here node server.mjs在 app.quanticdata.io 获取密钥 — 有免费月度额度且无需信用卡;当前价格请参阅 quanticdata.io 上的定价。
在客户端中安装
Claude Code
claude mcp add quanticdata \
-e QUANTICDATA_API_KEY=qd_live_your_key_here \
-- node /absolute/path/to/quanticdata-mcp-server/server.mjsClaude Desktop — claude_desktop_config.json
{
"mcpServers": {
"quanticdata": {
"command": "node",
"args": ["/absolute/path/to/quanticdata-mcp-server/server.mjs"],
"env": { "QUANTICDATA_API_KEY": "qd_live_your_key_here" }
}
}
}Cursor — .cursor/mcp.json,格式与上述相同。
任何支持 stdio MCP 的客户端均可使用:该服务器实现了 initialize、tools/list、tools/call 和 ping,这构成了仅工具型服务器的完整接口。
Related MCP server: Spider MCP Server
工具
工具 | 代理获得的内容 | 价格 |
| 一个 URL → Markdown / HTML / 文本、CSS 或 AI 提取、可选渲染 | $0.0002 |
| 自然搜索结果 + 相关搜索、4 个搜索引擎、18 个垂直领域 | 从 $0.0005 |
| 网站的所有 URL(来自站点地图 + 首页链接),含总数 | $0.0005 |
| 异步 BFS 爬取 → 每页 Markdown | $0.0003/页 |
| URL 的无 JS 视图与渲染视图对比,含差异,以及面向机器人的元信息 | $0.0012 |
| 目录:模式、示例、健康状态、您的单价 | 免费 |
| 在语义输入上运行一个 Collector | 从 $0.0004/行 |
| 一次运行及其交付的行 | 免费 |
一切按成功付费:失败的调用不产生任何费用。
无需客户端即可试用
node smoke-test.mjs # handshake + tools/list, no API key needed或者手动驱动 — 它在 stdin 上使用换行分隔的 JSON-RPC:
printf '%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"cli","version":"1"}}}' \
'{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
| node server.mjs效果良好的提示词
"绘制 quanticdata.io 的地图,然后抓取最近五篇博客文章并总结主题。"
"在美国和德国搜索
serp api pricing,告诉我前十名有何不同。""对我们的定价页面运行 SEO 审计 — 它是否可以在没有 JavaScript 的情况下被索引?"
"使用
list_collectors,然后用local_business_leads拉取奥斯汀的 40 家牙科诊所。"
最后一个值得内化的模式:先让代理阅读目录,然后选择收集器。发布模式正是为了让模型能够正确选择。
配置
变量 | 默认值 | 说明 |
| — | 必填;密钥格式类似 |
|
| 可指向其他位置以使用代理或暂存主机 |
来自 API 的错误以工具结果的形式返回,带有 isError: true,而非协议错误,因此模型可以读取消息并自行纠正,而不是让对话中断。
相关链接
MIT 许可。
Available Tools
9 toolscollector_run_statusAInspect
Fetch one Collector run and its delivered rows by run_id.
Inputs: runId (required).
crawlAInspect
Start an asynchronous BFS crawl from a seed URL; each page comes back as Markdown. Returns a job id immediately — poll with crawl_status. $0.0003 per page, unfetched budget refunded. Prefer map + batch when you can filter the URL list first.
Inputs: url (required), depth, limit, render, +4 more.
crawl_statusAInspect
Poll a crawl job: status, pagesCrawled, pagesQueued and the pages fetched so far.
Inputs: jobId (required).
list_collectorsAInspect
List the 31 ready-made Collectors: versioned scrapers you run with a semantic input (keyword + location, ASIN, domain, handle…) instead of URLs — web_search, keyword_ideas, amazon_search, amazon_product, ebay_search, aliexpress_search, google_maps_places, place_reviews, google_jobs, linkedin_jobs, indeed_jobs, google_news, google_shopping, product_offers, hotels, youtube_search, youtube_channel, reddit_posts, instagram_profile, tiktok_profile, tiktok_video, zillow_search, app_store_apps, google_play_apps, linkedin_profile, linkedin_company, company_profile, site_contacts, local_business_leads, search_images, search_videos. Each entry carries its input and output schema, examples, health and your price per delivered row. Free to call.
Inputs: no named parameters.
mapAInspect
Discover every URL of a site in one call: robots.txt sitemaps + /sitemap.xml (including nested indexes) + same-domain homepage links, de-duplicated. Returns the list plus a site-wide total and a per-section summary. Always cheaper than crawling to find out how big a site is. $0.0005.
Inputs: url (required), limit, search, country, +2 more.
run_collectorAInspect
Run one Collector with its semantic input. Short runs return the rows directly; long runs return a run_id to poll with collector_run_status. Billed per delivered row — zero rows costs zero.
Inputs: slug (required), input (required), async.
scrapeAInspect
Fetch one URL and return it as Markdown (default), HTML or plain text, through residential proxies with real-browser TLS fingerprints. Set render:true only when the content is genuinely absent from the raw HTML. Use extract for CSS-selector JSON, or ai_prompt to have the page turned into structured data by an LLM. $0.0002 per page; a failed fetch is free.
Inputs: url (required), mode, format, render, +5 more.
searchAInspect
Structured search results from Google, Bing, DuckDuckGo or Yandex: organic rows with position, title, link, snippet and sitelinks, plus related searches and any local/news/shopping blocks the page carried. Eighteen verticals via search_type. From $0.0005 per search.
Inputs: query (required), num, lang, page, +4 more.
seo_auditAInspect
Fetch a URL twice — as a no-JS bot and fully rendered — and return both views, the diff between them (contentOnlyInJs, canonicalMissingNoJs, titleChanged…) and the bot-facing meta (robots, OG, JSON-LD types). The fastest way to answer 'is this page indexable without JavaScript'. $0.0012.
Inputs: url (required), country, no_render.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 9 tools
Each tool targets a distinct job: single fetch, search, site discovery, crawl, crawl status, SEO audit, and collector management. Minor ambiguity exists between the `search` tool and the `web_search` collector, and between `scrape` and `seo_audit`, but the descriptions provide enough context to choose correctly.
Bare imperative verbs (`scrape`, `search`, `map`, `crawl`) sit alongside verb-noun names (`run_collector`, `list_collectors`) and noun-phrase status names (`crawl_status`, `collector_run_status`, `seo_audit`). The names are readable, but the pattern is not consistent enough to predict how a new tool would be named.
Nine tools is a well-scoped size for a data-acquisition server covering scraping, search, crawling, SEO auditing, and collector workflows. Each tool earns its place without overwhelming an agent.
The server covers scraping, search, URL discovery, crawling with status polling, SEO auditing, and collector execution/status, forming complete data-fetch workflows. Minor gaps include a referenced batch mode that is not exposed as a tool and no cancellation/list-jobs capability, but these can be worked around.
Maintenance
Related MCP Connectors
Web scraping, search, crawling, structured extraction and research for AI agents.
Web data for agents. Extract web pages, crawl websites, discover URLs, and search the web.
Web data for agents: YouTube transcripts, screenshots, Google News, WHOIS, jobs, tech stack, more.
Direct access to 60+ scraping and search tools. Extract structured data from Google (Search, Maps, Trends), Amazon, Airbnb, Social Media, and any web page directly into your AI agent.
Related MCP Servers
- AlicenseAqualityCmaintenanceEnables AI assistants to access real-time web data through search, markdown scraping, and browser automation while bypassing anti-bot protections. It provides tools for web research, e-commerce monitoring, and data extraction from across the globe.45MIT
- AlicenseAqualityCmaintenanceEnables AI agents to crawl, scrape, search, and automate browsers with anti-bot bypass, providing fast web access via 22 tools.2259 npm3MIT
- 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.-
- AlicenseAqualityBmaintenanceEnables AI agents to perform web searches, fetch and extract page content, and crawl sites with caching, rate limiting, and robots.txt compliance, all without needing API keys.171,855 PyPI2MIT