M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
S

ScrapeGraph MCP Server

Community
AI-powered web scraping and data extraction with enterprise reliability via ScrapeGraph API.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #227 of 438
Stars
★ 115 Popular
Transport
stdio (local process)
Runtime
Python 3.13+ · Node.js · Hosted (no local runtime)
Credentials
API key / credential required
License
MIT
Last commit
Tools
17
52FMRS · D

This MCP server offers AI-powered web scraping functionality with a rich set of tools covering scrape, extract, search, crawl, and monitor. Setup is easy but requires a paid API key. Note that tool names were renamed in v3; consult the documentation when upgrading.

Strongest · Documentation 14/20 Weakest · Reliability 5/20

Reliability
5/20
Security and permissions
9/20
Maintenance
12/20
Documentation
14/20
Setup experience
12/20
Why each score
Reliability 5/20
Only the README and manifest were supplied; no actual repository source files, CI workflows, or committed tests were available for review. The described tool set (scrape, extract, search, crawl_start/crawl_get_status, schema, credits, history, monitors) is internally consistent, but real behavior cannot be verified, and error/edge-case handling is claimed rather than demonstrated. The manifest also marks SGAI_API_KEY optional while the README implies it is required, and this ambiguity is unverified. Deducted heavily for the absence of executable evidence; score kept below 10.
Security and permissions 9/20
No red-line indicators were found: no malware, no real secrets in examples, and no destructive default operations. Credential handling is conventional (SGAI_API_KEY via environment or MCP config, marked secret in the manifest). However, without source code, least-privilege boundaries, data-flow disclosure, and confirmation for long-running crawl/monitor operations cannot be verified; the manifest lists the API key as optional while the README treats it as required, and cost/quota guardrails are not specified. Deducted for missing source-level security evidence.
Maintenance 12/20
The repository is not archived, has an MIT license, a concrete version (3.0.0), and the README documents a v2-to-v3 migration with a tool renaming table, which suggests active stewardship. But no commit history, release dates, issue-response data, or dependency-update evidence was supplied, so sustained maintenance cannot be confirmed. Stars and issue counts are discovery signals only and were not used to add points; scored conservatively at 12.
Documentation 14/20
The README provides layered guidance: quick start, authentication, tool tables with output formats and parameter limits (e.g., num_results clamped 3–20), local/remote/Smithery/ADK configuration examples, troubleshooting, and common error handling. Deductions: examples are not source-verified, no explicit cost/plan disclosure, inconsistent spelling (ScapeGraph vs ScrapeGraph), and the remote endpoint is presented as a migration warning rather than stable documentation.
Setup experience 12/20
Multiple short install paths are documented: Smithery via npx, pip/uv local install, remote HTTP, plus Claude Desktop and Cursor configuration examples with Windows-specific notes. However, static calibration caps setup at 15 without real CI artifacts and committed tests proving that these commands produce a working connection; no such evidence was supplied, and the npx/Smithery path adds external-service fragility. Deducted to 12.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkChanges third-party account data

Best for

  • AI assistant developers who need natural-language-driven web scraping and structured data extraction.
  • Crawler and data analysis workflows that need automated extraction from multiple pages.
  • Users who need to monitor website changes and extract information on a schedule.

Not for

  • Developers who don't need AI-assisted scraping and just want simple direct HTTP requests.
  • Users without a ScrapeGraph AI paid subscription or with limited credits, as API calls may incur costs.
  • Users needing local JavaScript rendering without using ScrapeGraph API's rendering capabilities.

Required permissions

  • Requires a ScrapeGraph API key (SGAI_API_KEY) to access ScrapeGraph services.
  • Requires network access to reach the ScrapeGraph API (default https://v2-api.scrapegraphai.com/api).
  • API key passed via environment variable or MCP config, which may involve secret management.

Risks and side effects

  • API key leakage could lead to unauthorized usage and credit loss.
  • API calls are subject to rate limiting and quota management; exceeding may incur extra costs.
  • Scraping websites may violate their terms of service or laws; use responsibly.

Setup

Before you start

Runtime：Python 3.13+ · Node.js · Hosted (no local runtime)

SGAI_API_KEY requiredsecret Required ScrapeGraph API key; obtain it from dashboard.scrapegraphai.com after signing up. Can also be passed via MCP config scrapegraphApiKey or the X-API-Key header.
Other optional settings (2)
SGAI_API_URL optional Optional; overrides the API base URL (default https://v2-api.scrapegraphai.com/api).
SGAI_TIMEOUT optional Optional; request timeout in seconds (default 120).
  1. Obtain an API key from the ScrapeGraph Dashboard.
  2. Quick install with Smithery: npx -y @smithery/cli install @ScrapeGraphAI/scrapegraph-mcp --client claude.
  3. Or install locally: clone repo, pip install -e ., then set environment variable SGAI_API_KEY and run with scrapegraph-mcp or python -m scrapegraph_mcp.server.
  4. Configure Claude Desktop or Cursor using the JSON configs provided in the README.
claude_desktop_config.json 
{
  "mcpServers": {
    "scrapegraph-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli@latest",
        "run",
        "@ScrapeGraphAI/scrapegraph-mcp",
        "--config",
        "\"{\\\"scrapegraphApiKey\\\":\\\"YOUR-SGAI-API-KEY\\\"}\""
      ]
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "scrapegraph-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli@latest",
        "run",
        "@ScrapeGraphAI/scrapegraph-mcp",
        "--config",
        "\"{\\\"scrapegraphApiKey\\\":\\\"YOUR-SGAI-API-KEY\\\"}\""
      ]
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add scrapegraph-mcp -- npx -y @smithery/cli@latest run @ScrapeGraphAI/scrapegraph-mcp --config '"{\"scrapegraphApiKey\":\"YOUR-SGAI-API-KEY\"}"'

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

The client's tool list should show tools like scrape, extract, and search; asking 'Show my ScrapeGraph remaining credits' should return a number via the credits tool, proving the connection works.

Troubleshooting

  1. If tools don't appear in client, check that the server starts properly and the API key is correct.
  2. If you get 401 Unauthorized, verify your API key is valid.
  3. If you get 402 Payment Required, check your account credits.
  4. If crawl doesn't return results, keep polling crawl_get_status until status is 'completed'.
  5. On Windows, use the specific command or check PATH settings.

Things to try

Once connected, you can ask your AI assistant things like:

  • Convert https://scrapegraphai.com to markdown
  • Extract all product names, prices, and ratings from this e-commerce page
  • Research and summarize recent developments in AI-powered web scraping
  • Run this extract prompt on https://example.com every day at 9am

Tools 17

scrape read-only
Scrape a webpage and output multiple formats: markdown, html, screenshot, branding, links, images, summary.
extract read-only
Extract structured data based on a prompt, requiring website_url and user_prompt, with optional output_schema.
search read-only
Perform a search returning up to 20 results, supporting country_search, time_range, and output_schema.
crawl_start writes
Start async multi-page crawl with extraction_mode: markdown, html, links, images, summary, branding, screenshot.
crawl_get_status read-only
Get status of a crawl job; poll until status is completed.
crawl_stop writes
Stop an ongoing crawl job.
crawl_resume writes
Resume a paused crawl job.
schema read-only
Generate or augment a JSON Schema from a prompt.
Show 9 more tools
credits read-only
Get account credits.
history read-only
Get paginated usage history, optionally filtered by service.
monitor_create writes
Create a scheduled monitoring job.
monitor_list read-only
List all monitoring jobs.
monitor_get read-only
Get details of a specific monitoring job.
monitor_pause writes
Pause a monitoring job.
monitor_resume writes
Resume a monitoring job.
monitor_delete destructive
Delete a monitoring job.
monitor_activity read-only
Get paginated activity history of a monitoring job (id, createdAt, status, changed, elapsedMs, diffs).

Use cases

Convert a webpage to Markdown for easy LLM processing.
Extract all product names, prices, and ratings from an e-commerce page.
Search for recent AI developments and get a structured summary.
Crawl multiple pages of a site to map content or capture API endpoints.
Set up daily scheduled monitoring to track changes on specific data.

Supported clients

Claude Desktop
Cursor

Listed from the project's documentation, not tested by this site.

Overview

The ScrapeGraph MCP Server is a production-ready Model Context Protocol (MCP) server that provides seamless integration with the ScrapeGraph AI API. It enables language models to leverage advanced AI-powered web scraping capabilities, offering tools for scraping (multi-format), prompt-based extraction, search, async multi-page crawling (start/status/stop/resume), JSON schema generation, scheduled monitoring (create/list/get/pause/resume/delete/activity), account credits, and history. The server uses stdio transport and can be configured via environment variables (SGAI_API_KEY, SGAI_API_URL, SGAI_TIMEOUT). Note that v3 has renamed tools, e.g., smartscraper to extract, smartcrawler_initiate to crawl_start.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 8372895aa6ea Data synced 2026-10-12 Read the FMRS scoring method