MCP Server for Cursor — Web Data in Your Development Workflow

Give Cursor's AI agent the ability to scrape web pages, capture screenshots, extract metadata, and query any URL. Install in under 2 minutes — no code changes required.

No credit card required.

Cursor IDE + MCP

Cursor IDEAI-powered code editor

stdio

opengraph-io-mcpruns locally on your machine

REST API

OpenGraph.io APIapi.opengraph.io

Tools Available in Cursor

Link Previews

Screenshots

Web Scraping

AI Query

npx -y opengraph-io-mcp

Developer Workflow

Supercharge Your Cursor AI Agent with Web Data

Cursor's AI agent becomes dramatically more useful when it can access live web data. These tools are discovered automatically — just ask in natural language.

Extract Metadata

Fetch OpenGraph tags, Twitter Cards, and meta tags from any URL — debug your own site's link previews or audit competitors without leaving your editor.

get_og_data

Capture Screenshots

Take full-page or viewport-sized screenshots of any webpage. Compare competitor UIs, verify deployments, or generate visual documentation inline.

screenshot

Scrape Web Pages

Fetch rendered HTML with JavaScript execution and bot detection bypass. Pull API docs, extract structured data, or analyze page content mid-coding session.

scrape

Query Any Page

Ask natural-language questions about any webpage and get structured answers. Extract pricing tables, summarize documentation, or validate page content.

query

Setup Guide

Set Up Cursor MCP in Under 2 Minutes

Add the OpenGraph.io MCP server to Cursor with a single JSON config. No extensions to install, no build steps — just paste and restart.

1Get your free App ID: Sign up at dashboard.opengraph.io — the free tier includes 100 credits per month.
2Create or edit mcp.json: Use ~/.cursor/mcp.json for global access or .cursor/mcp.json in your project root.
3Paste the config: Add the OpenGraph.io MCP server entry with your App ID (see the config panel).
4Restart Cursor: Fully restart (not just reload). Verify by asking “What tools do you have access to?” in chat.

~/.cursor/mcp.json

{
  "mcpServers": {
    "opengraph": {
      "command": "npx",
      "args": [
        "-y",
        "opengraph-io-mcp"
      ],
      "env": {
        "OPENGRAPH_APP_ID": "YOUR_OPENGRAPH_APP_ID"
      }
    }
  }
}

Available across all Cursor workspaces.

Config Scopes

Global vs. Project-Level MCP Config

Cursor supports two config locations. Choose based on whether you need tools everywhere or just for a specific project.

Global (~/.cursor/mcp.json)

Available in every Cursor workspace
Great for personal tools like web data access
Set up once, use everywhere
Lives in your home directory

Project (.cursor/mcp.json)

Scoped to a single project
Commit to git for team-wide access
Required for Cloud Agents
Takes priority over global config

Developer Prompts

Real-World Prompts for Cursor

The AI agent automatically picks the right MCP tool based on your request. These prompts work in Cursor's chat, Composer, or inline editing.

API research: “Scrape the Stripe API docs at stripe.com/docs/api and show me how to create a payment intent.”
UI analysis: “Screenshot linear.app and describe their hero section design, then help me build something similar.”
OG tag debugging: “Unfurl my staging site at staging.myapp.com and check if the OpenGraph tags are set up correctly.”
Data extraction: “Scrape news.ycombinator.com and extract the top 10 story headlines with their URLs into a JSON array.”
Competitive analysis: “Query competitor.com/pricing and create a comparison table of their plan features vs ours.”

API & Docs Research

Scrape documentation pages and have the agent summarize endpoints, parameters, and code examples inline.

Design & UI Reference

Screenshot competitor pages, analyze their design patterns, and generate implementation code — all in one flow.

SEO & Link Preview QA

Test your OG tags, validate metadata, and debug social sharing previews without switching to a browser.

Security

Your API Key Never Reaches the AI Model

The MCP server runs locally as a subprocess of Cursor. Your App ID stays in your config file and is never sent to the AI model.

Local Config Storage

Your App ID is stored in mcp.json on your machine — never exposed to the AI model or transmitted externally.

Local Subprocess

The MCP server runs as a child process of Cursor on your machine. No cloud hosting, no shared infrastructure.

Stdio Transport

Communication uses standard input/output between Cursor and the MCP server. No open ports, no network exposure.

No Persistent State

The server doesn't cache responses locally or maintain session state. Every request is stateless and ephemeral.

Troubleshooting

Common Issues & Fixes

Most issues are solved by one of these steps. If you're still stuck, check the full troubleshooting guide.

Tools not appearing: Fully restart Cursor (not just reload the window). Then check Settings > Tools & MCP to verify the server is listed and enabled.
JSON parse error: Validate your mcp.json at jsonlint.com. Watch for trailing commas and duplicate keys.
Node.js not found: Ensure Node.js 18+ is installed — run node --version in your terminal. Restart Cursor after updating your PATH.
View MCP logs: Open the Output panel (Cmd+Shift+U) and select “MCP Logs” from the dropdown for detailed error messages.

Cursor MCP Shortcuts

Settings: Cmd+Shift+J > Tools & MCP

Logs: Cmd+Shift+U > MCP Logs

Verify: “What tools do you have?”

Cross-Platform

One MCP Server, 8+ AI Clients

The same opengraph-io-mcp package powers every client below. Only the config file format differs.

Zed

Our Clients

Trusted by Thousands

Over 1B Site Previews!

Try it for free today

Get started with 100 free API requests per month. No credit card required. Experience the power of our API with zero commitment.

FAQ

Create or edit the file ~/.cursor/mcp.json (for global access) or .cursor/mcp.json in your project root (for project-specific access). Add the OpenGraph.io MCP server config with your App ID, save the file, and restart Cursor. The AI agent will automatically discover the available tools.

Cursor supports the Model Context Protocol (MCP), an open standard that connects Cursor's AI agent to external tools and data sources. The OpenGraph.io MCP server runs locally on your machine and gives the agent access to web scraping, metadata extraction, screenshot capture, and AI-powered page queries — all callable directly from Cursor's chat or Composer.

Global config (~/.cursor/mcp.json) makes MCP tools available across all your Cursor workspaces. Project-level config (.cursor/mcp.json) only applies to that project and can be committed to git so your team gets the same tools. Both files are merged — if the same server name appears in both, the project-level config takes priority. Cloud Agents only pick up project-level configs.

Cursor's agent gains four web data tools: (1) Extract OpenGraph metadata, Twitter Cards, and HTML meta tags from any URL, (2) Capture full-page screenshots at any viewport size, (3) Scrape rendered HTML with JavaScript execution and bot detection bypass, and (4) Ask natural-language questions about any webpage and get structured answers — all directly in your development workflow.

The MCP server itself is free and open-source. It connects to the OpenGraph.io API, which includes a free tier with 100 credits per month. Metadata extraction costs 1 credit, screenshots cost 20 credits (1 if cached), and AI queries cost 100–200 credits. Paid plans start at $25/month for 50,000 credits.

First, fully restart Cursor (not just reload the window). Then verify your mcp.json has valid JSON syntax with no trailing commas. Check that Node.js 18+ is installed by running node --version in your terminal. You can view MCP-specific logs by opening the Output panel (Cmd+Shift+U on Mac, Ctrl+Shift+U on Windows/Linux) and selecting 'MCP Logs'. You can also toggle servers in Settings > Tools & MCP.

Yes. With the OpenGraph.io MCP server installed, just ask Cursor's AI agent in natural language: 'Scrape the pricing page at example.com and create a comparison table.' The agent calls the scrape tool automatically, fetches the rendered HTML, and processes it — no HTTP requests, API wrappers, or boilerplate code needed.

Yes. Cloud Agents pick up MCP servers from your project's .cursor/mcp.json file. Make sure the config is in the project directory (not the global ~/.cursor/mcp.json) so cloud agents can find and use the tools.

Still have Questions? 🤔

Can't find the answer you're looking for? Please chat to our friendly team. We usually respond within 24 hours.