> ## Agent Instructions
>
> Base URL: https://api.anysite.io
> Authentication: send the `access-token` header. Do NOT use `Authorization: Bearer`.
> Full endpoint catalog: https://app.anysite.io/docs
# MCP Server Overview

Anysite MCP Server — 5 meta-tools accessing 3,600+ endpoints across 600+ data sources

## What is Anysite MCP Server?

The Anysite MCP Server gives AI agents (Claude Desktop, Claude Code, Cursor, n8n) direct access to **3,600+ data endpoints** across **600+ sources** — LinkedIn, Instagram, Twitter, Reddit, YouTube, ProductHunt, Google, SEC, and many more.

Instead of calling REST APIs manually, your AI agent discovers available endpoints, fetches data, and analyzes results — all through a single MCP connection.

  
#### 3,600+ Endpoints

    LinkedIn, Instagram, Twitter, Reddit, YouTube, ProductHunt, Google, SEC, and 30+ more sources
  

  
#### 5 Meta-Tools

    discover → execute → get_page → query_cache → export_data
  

  
#### Zero Config

    New endpoints appear automatically — no server updates needed
  

---

## Architecture: 5 Meta-Tools

> 
The Anysite MCP Server uses a **meta-tool architecture** with 5 universal tools instead of individual tools per endpoint. This reduces context window usage by 2.5x and supports dynamic endpoint discovery.

| Tool | Purpose |
|------|---------|
| **discover** | Learn what endpoints exist and what parameters they accept |
| **execute** | Fetch data from any endpoint |
| **get_page** | Paginate through large result sets from cache |
| **query_cache** | Filter, sort, and aggregate cached results server-side |
| **export_data** | Download full datasets as JSON/CSV/JSONL files |

### Workflow

```text
1. discover("linkedin", "user")
   → Returns endpoint names, parameter schemas, and LLM hints

2. execute("linkedin", "user", "get", {"user": "darioamodei"})
   → Fetches data, returns first 10 items + cache_key

3. get_page(cache_key, offset=10)
   → Next page from cache (no new API call)

4. query_cache(cache_key, conditions=[{"field": "followers", "op": ">", "value": 1000}])
   → Filter/aggregate in cache without loading data into context

5. export_data(cache_key, "csv")
   → Download full dataset as file
```

> 
**Why meta-tools?** Instead of loading 1,100+ tool definitions into the LLM context, only 5 compact definitions are needed. The LLM first discovers what's available via `discover()`, then calls the right endpoint with correct parameters via `execute()`.

---

## Data Sources

  
#### LinkedIn

    **35 endpoints**

    Profile search, company data, posts, comments, skills, experience, and more
  

  
#### Instagram

    **8 endpoints**

    Posts, comments, likes, user profiles, friendships, and reels
  

  
#### Twitter/X

    **4 endpoints**

    User search, profiles, posts, and content extraction
  

  
#### Reddit

    **11 endpoints**

    Post search, user posts, comments, and content extraction
  

  
#### YouTube

    **14 endpoints**

    Video search, channel videos, video details, comments, and subtitles
  

  
#### ProductHunt

    **23 endpoints**

    Products, launches, makers, topics, and collections
  

  
#### Google

    **10 endpoints**

    Web search, news, maps, and related results
  

  
#### Web Parser

    **2 endpoints**

    Webpage parsing and sitemap extraction
  

  
#### Y Combinator

    **3 endpoints**

    YC company data, founder search, and startup intelligence
  

  
#### SEC

    **2 endpoints**

    SEC company filings and document search
  

  
#### DuckDuckGo

    **1 endpoint**

    Web search powered by DuckDuckGo
  

  
#### Other Sources

    **~70 endpoints**

    Amazon, Facebook, TikTok, Yahoo, Crunchbase, Companies House, GitHub, and 20+ more
  

#### View All Endpoints

  Browse complete reference of all 1,100+ available MCP endpoints

---

## Quick Start

  
#### Get your MCP URL

    Sign up at [app.anysite.io](https://app.anysite.io) and copy your Remote MCP Server URL
  

  
#### Connect your AI client

    Add the MCP URL to your preferred client:

    
      
#### Claude Desktop

        OAuth connection in 60 seconds
      
      
#### Claude Code

        CLI integration for developers
      
      
#### Cursor

        IDE integration for coding
      
      
#### n8n

        Workflow automation
      
    
  

  
#### Start using

    Ask your AI agent naturally — it will use `discover` and `execute` automatically:

    > "Find the LinkedIn profile of Sam Altman and show me his recent posts"

    > "Search for YC companies in the AI space from batch W2024"
  

---

## Next Steps

  
#### All Endpoints

    Full reference of 3,600+ endpoints with parameters and examples
  

  
#### Data Analysis

    Server-side filtering, aggregation, and export with query_cache
  

  
#### Custom Endpoints

    Build your own endpoints that appear automatically in MCP
  

  
#### Unlimited Plan

    $30/month for unlimited MCP requests