The server is open source under an MIT license and free to run, but it calls the ScrapeGraphAI API, which is paid and metered in credits against your SGAI_API_KEY. The credits tool reports your remaining balance.

What output formats can scrape return?

scrape can return markdown, HTML, a screenshot, branding data, extracted links, images, or a summary, selected via the output_format parameter; extract returns structured JSON against an optional schema.

ScrapeGraphAI MCP server

OfficialScrapeGraphAIConfig last verified Jun 1, 2026

ScrapeGraphAI's official MCP server: AI-powered scraping, structured extraction, web search, multi-page crawling, and scheduled page-change monitoring.

ScrapeGraphAI MCP is the official server from ScrapeGraphAI, the AI-powered web-data API. It gives an agent a full extraction toolbox: scrape a single page into markdown, HTML, links, images, or a screenshot; extract typed structured data against a prompt and optional JSON Schema; search the web and get structured results; and crawl across many pages in markdown or AI mode. Crawls run asynchronously with start, status, stop, and resume controls, and a schema tool can generate or augment a JSON Schema from a plain-language prompt so the agent knows the exact shape to ask for. A set of monitor tools schedule recurring scrapes that track page changes over time, and credits and history tools let the agent watch its own usage.

It installs two ways. Locally it runs over stdio and is published to PyPI, so the simplest path is uvx scrapegraph-mcp with an SGAI_API_KEY environment variable; optional variables override the API base URL and request timeout. ScrapeGraphAI also hosts a remote endpoint at https://scrapegraph-mcp.onrender.com/mcp, which authenticates with your API key supplied as an X-API-Key header. The API key meters credits against your ScrapeGraphAI account either way.

Quick install

Copy-paste configs are provided for all 8 supported clients. Pick your client below.

Add to ~/.claude.json

~/.claude.json

json

{
  "mcpServers": {
    "scrapegraph": {
      "command": "uvx",
      "args": [
        "scrapegraph-mcp"
      ],
      "env": {
        "SGAI_API_KEY": "<SGAI_API_KEY>",
        "SGAI_API_URL": "<SGAI_API_URL>",
        "SGAI_TIMEOUT": "<SGAI_TIMEOUT>"
      }
    }
  }
}

Or via CLI

bash

claude mcp add scrapegraph -- uvx scrapegraph-mcp

Step-by-step guides

Available tools

Tool	Description
scrape	Scrape a single URL, returning markdown, HTML, screenshot, branding, links, images, or a summary.
extract	Extract structured data from a page given a website_url and user_prompt, with an optional output_schema.
search	Search the web and return structured results, with optional country, time range, and output schema.
crawl_start	Start an asynchronous multi-page crawl in markdown, html, links, images, summary, branding, or screenshot mode.
crawl_get_status	Poll a crawl job by id until its status is completed and retrieve the results.
crawl_stop	Stop an active crawl operation.
crawl_resume	Resume a paused crawl operation.
schema	Generate or augment a JSON Schema from a plain-language prompt for use in extraction.
credits	Get the remaining credit balance for the account.
history	Retrieve paginated request history, optionally filtered by service.
monitor_create	Create a scheduled scraping job that tracks page changes over time.
monitor_list	List all configured monitors.
monitor_get	Get the details of a single monitor.
monitor_pause	Pause a monitor.
monitor_resume	Resume a paused monitor.
monitor_delete	Delete a monitor.
monitor_activity	Retrieve a monitor's paginated tick history of checks.

Required configuration

SGAI_API_KEYRequired
ScrapeGraphAI API key used to authenticate requests and meter credits. Required. Can also be passed via the X-API-Key header on the remote endpoint.
SGAI_API_URLOptional
Override the API base URL. Optional; defaults to https://v2-api.scrapegraphai.com/api.
SGAI_TIMEOUTOptional
Per-request timeout in seconds. Optional; defaults to 120.

What you can do with it

Extract typed data from any page

Give the agent a URL and a description of the fields you want; it calls extract with an output_schema (which schema can generate for you) and gets back clean structured JSON instead of raw HTML.

Crawl a site and watch it for changes

Kick off crawl_start to gather a subtree of pages, poll crawl_get_status until it completes, then stand up a monitor with monitor_create so the agent is alerted when a tracked page changes.

FAQ

Is it free?: The server is open source under an MIT license and free to run, but it calls the ScrapeGraphAI API, which is paid and metered in credits against your SGAI_API_KEY. The credits tool reports your remaining balance.
Does it support remote/OAuth?: It supports both. Locally it runs over stdio via uvx scrapegraph-mcp with an SGAI_API_KEY. ScrapeGraphAI also hosts a remote endpoint at https://scrapegraph-mcp.onrender.com/mcp that authenticates with the API key as an X-API-Key header rather than OAuth.
What output formats can scrape return?: scrape can return markdown, HTML, a screenshot, branding data, extracted links, images, or a summary, selected via the output_format parameter; extract returns structured JSON against an optional schema.

← Browse all search-and-data servers