15 个工具 · 7 个可使用平台 key · 8 个仅自有 key
ScrapeGraphAI API：15 个工具，按次付费
通过 Toolswitch 使用 ScrapeGraphAI 工具。
常见任务：web.extract, web.search, web.crawl。
能做什么
Fetch owned crawl pages with each page's resolved scrape result.
Read an owned crawl's state and lightweight page metadata.
Fetch one public HTML page and return clean Markdown.
在 Agent 里使用
通过 MCP 将这些工具提供给 Agent，或使用 HTTP 将调用接入你的工作流。
- MCP: 通过 MCP 连接 Agent，以发现和使用工具。
- HTTP: 通过 HTTP API 在应用中调用工具。
为什么通过 Toolswitch 调用？
无需供应商账号
使用平台 key 的工具无需单独注册供应商账号。
按次付费
按实际使用量付费，无需供应商订阅。
价格透明
调用前即可查看工具价格与计费单位。
全部 15 个工具
Open web & SEO data · 15 个工具
Fetch owned crawl pages with each page's resolved scrape result.
scrapegraphai.web.crawl.pagesRead an owned crawl's state and lightweight page metadata.
scrapegraphai.web.crawl.statusRead recent ticks, completion states and per-format change diffs.
scrapegraphai.web.monitor.activityPermanently remove a monitor and its stored tick history.
scrapegraphai.web.monitor.deleteRead one monitor's status, schedule and capture configuration.
scrapegraphai.web.monitor.getList every scheduled monitor on the connected account.
scrapegraphai.web.monitor.listStop future ticks while preserving the monitor and its activity.
scrapegraphai.web.monitor.pausePut a paused monitor back on its saved schedule.
scrapegraphai.web.monitor.resumeChange a monitor's URL, name, schedule, formats, webhook or fetch configuration.
scrapegraphai.web.monitor.updateFetch one public HTML page and return clean Markdown.
scrapegraphai.web.scrapeStart a bounded one-page Markdown crawl and return its job id.
scrapegraphai.web.crawlFetch one public HTML page and use a natural-language prompt to return structured JSON.
scrapegraphai.web.extractSchedule recurring page captures and change detection on the connected account.
scrapegraphai.web.monitor.createSearch the web and return the top HTML pages with Markdown content inline.
scrapegraphai.web.searchSearch HTML pages, return their Markdown, and run one AI extraction across the results.
scrapegraphai.web.search.extract
限制与计费
The shared account is configured for 500 requests/minute, one concurrent crawl and one monitor. Live responses returned no usable X-RateLimit-* or Retry-After headers, so treg enforces the supplied 500 requests/minute ceiling directly. Platform extraction and search accept HTML only, and platform Crawl is fixed to one Markdown page so every hold is bounded. Monitor tools require the team's own key because scheduled ticks spend credits outside a foreground treg call.
官方定价 ↗