The MCP Server for Live Web Data
Give Claude, Cursor, or ChatGPT live access to Instagram, Twitter/X, Reddit, YouTube, business data, and any URL. Ask in plain English. Typed JSON comes back.
Why Teams Pick It
Natural-Language Queries
Ask in plain English — “Find VPs of sales at SaaS companies in New York.” No endpoint lookup, no parameters to learn, no code to write.
Every Major Platform + Any URL
Ready-made endpoints for Instagram, Twitter/X, Reddit, YouTube, Google Maps, SEC EDGAR, and Crunchbase. Point the web parser at any other URL and typed JSON comes back. The catalog is dynamic — new sources appear automatically, with nothing to reinstall.
Business Data Included
Coverage reaches the business data page readers stop at — profiles, company pages, posts, and people search from behind the login wall.
Server-Side Analysis
Filter, sort, and aggregate cached results on the server. Your agent works with targeted fields, not raw dumps — context stays clean and token costs stay low. Results sit in a ClickHouse cache with 7-day export, and leave as JSON, CSV, or JSONL.
Flat Plans, No Meter
From $30/month flat. Every plan covers all five tools and every source, and includes the Person Analyzer and Competitor Analyzer skills — no credits to count, no per-request meter. Upgrade only for more headroom.
Works in Your Tools
Claude Desktop connects over OAuth — no keys to manage. Claude Code, Cursor, and any MCP 1.0 client take an API key in a single config line — including n8n, when no-code automation is your interface.
How It Works
Install Once
Add the MCP server to Claude Desktop, Claude Code, or Cursor. One connection string, about 60 seconds.
Ask Naturally
Describe the data you need the way you’d brief a teammate — no endpoints to memorize.
“Find marketing managers in San Francisco with agency experience.”
Get Typed JSON
Named fields land directly in the conversation — people, companies, posts — not markup to untangle.
{ "name": "…", "headline": "…", "location": "San Francisco, CA" }
Act on Results
Your agent analyzes, filters, and exports right there. No context switching, no copy-pasting.
Connect in 60 Seconds
https://mcp.anysite.io/mcp
The OAuth URL. In Claude Desktop: Settings → Connectors → Add Custom Connector, paste it, click Connect, authorize once in your browser — no API key involved.
claude mcp add --transport http anysite \ "https://mcp.anysite.io/mcp?api_key=YOUR_KEY"
One line in your terminal. claude mcp list then confirms the server is connected.
{
"mcpServers": {
"anysite": {
"url": "https://mcp.anysite.io/mcp?api_key=YOUR_KEY",
"transport": "http"
}
}
}Goes in ~/.cursor/config.json. Type /mcp in Cursor to see the tools appear.
What You Ask. What You Get.
“Find CTOs at Series A fintech startups in New York with an AI or ML background.”
What happens: The server discovers the right search endpoint, runs the query, caches results, and filters by seniority server-side — your agent gets a targeted list, not 200 raw profiles.
You get: A typed list of CTO profiles — name, company, current role, profile link, and email where available.
“Pull our main competitor’s last 50 posts across social. Summarize product announcements and engagement.”
What happens: Fetches company posts from the professional network and Twitter/X in parallel, then aggregates average engagement by post type server-side before your agent summarizes.
You get: Announcements grouped by topic, average engagement by post type, and notable spikes flagged.
“Search Reddit and Hacker News for discussions about our category from the last 30 days. What are the top pain points?”
What happens: Searches both communities with time filters, then narrows by comment count and sorts by recency server-side — your agent extracts pain points from an already-targeted set.
You get: Recurring pain points ranked by frequency, with the source threads linked.
Common Questions
Every MCP plan includes usage limits that keep the service fast and reliable for everyone. Your plan comes with two windows that apply at the same time: a 5-hour limit for any rolling 5-hour period — the one you may notice during long, intensive sessions — and a weekly limit for any rolling 7-day period, sized so regular daily use fits comfortably. Some tool categories (for example, business data tools) have their own separate 5-hour and weekly limits in addition to the overall ones.
Usage is based on work, not request count: lightweight calls use very little of your limit, heavy data-extraction requests use more — so two users making the same number of requests can consume their limits at very different rates.
Limits are rolling, with no fixed reset time. Each window slides: usage older than 5 hours (or 7 days) automatically stops counting against you, so capacity frees up gradually as earlier activity ages out — you never wait for a calendar day or week to end. All API keys and MCP sessions on the same account draw from the same limits; creating an additional key does not grant additional capacity.
If you hit a limit, requests are temporarily declined with a rate-limit error and your MCP client will tell you the limit was reached. Nothing is lost — wait a bit and continue once the window frees up; your credit balance is unaffected, limits and credits are independent. You can see your current usage for each window as a percentage in your dashboard. Need more headroom? Higher-tier plans come with higher limits, and upgrading applies the new limits immediately.
Get Web Data in Your Next Conversation
Connect in about a minute, ask in plain English, and typed JSON from 550+ sources lands in your chat — from $30/month with a 7-day free trial.