ScrapeGraphAI scrapegraph MCP server

What it does

An MCP server that integrates with the ScrapeGraph AI API to enable language models to perform web scraping, data extraction, searching, and crawling. For developers who want to add AI-powered web scraping capabilities to their AI assistants or applications.

• Extract structured data from web pages using natural language prompts without writing CSS selectors or XPath queries.
• Search the web and retrieve results in a specified format or schema, with filtering by country and time range.
• Crawl multiple pages of a website asynchronously to gather data in markdown, HTML, links, images, or other formats.

Tools this server exposes

Tool	Does
scrape	Fetch a URL and convert it to markdown, HTML, screenshot, branding data, links, images, or summary.
extract	Extract data from a website URL using a natural language prompt and optional JSON schema.
search	Search the web with optional filters for country and time range, returning results in a specified schema.
crawl_start	Begin an asynchronous crawl of multiple pages with a specified extraction mode.
crawl_get_status	Check the status and results of an ongoing or completed crawl job.
crawl_stop	Stop an active crawl job.
crawl_resume	Resume a paused crawl job.
schema	Generate or augment a JSON schema from a natural language prompt.
credits	Check your remaining API credits.
history	Retrieve paginated history of past API requests, with optional filtering by service type.
monitor_create	Create a scheduled job to monitor a URL at regular intervals.
monitor_list	List all scheduled monitoring jobs.
monitor_get	Get details of a specific monitoring job.
monitor_pause	Pause a scheduled monitoring job.
monitor_resume	Resume a paused monitoring job.
monitor_delete	Delete a scheduled monitoring job.
monitor_activity	Retrieve paginated history of execution results for a monitoring job, including status and changes detected.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.