anysite

Anysite MCP Server provides a centralized gateway for AI agents to access web data via MCP (Model Context Protocol). It exposes a suite of tools that cover social platforms (LinkedIn, Instagram, Reddit, Twitter) as well as generic web parsing, enabling agents to search profiles, fetch posts, extract metadata, and monitor mentions in real time. The server is designed with an agent-first approach, returning structured data that agents can consume directly, and it includes OAuth-based authentication to simplify secure connections for clients like Claude Desktop, Cursor, Windsurf, and other MCP-enabled tools. Typical workflows involve querying for people, posts, or domains and then chaining results across platforms (e.g., find a LinkedIn profile, retrieve recent activity, then cross-check with a related Twitter or Instagram presence).

To use it, install or run the MCP server client, configure your MCP client with the provided mcpServers entry, and connect using the OAuth URL or your local credentials. Once connected, your agent can invoke a growing catalog of tools such as search_linkedin_users, get_linkedin_profile, get_instagram_user, search_reddit_posts, duckduckgo_search, search_youtube_videos, parse_webpage, and many more (the catalog lists 65+ tools). The server handles real-time data retrieval, structured responses, and self-healing behavior to adapt to platform changes and rate limits.

Prerequisites:

• Node.js and npm installed on your machine (recommended Node.js LTS).
• Git for cloning repositories if you choose the local development path.
• Access to the internet to fetch MCP tools from npm.

Option A: Quick start using remote MCP (recommended)

1. Ensure you have Node.js and npm installed.
2. Run the MCP server via npx (no local install required):

npx -y @anysite/mcp

This uses the prebuilt MCP package from npm and starts the server in your environment. The server will expose MCP endpoints that your MCP clients can consume.

Option B: Local MCP server (for development or customization)

1. Clone the repository and install dependencies:

# Clone the repository
git clone https://github.com/anysiteio/anysite-mcp-server.git
cd anysite-mcp-server

# Install dependencies
npm install

# Optional: build if there is a build step
npm run build

2. Create a local .env file with credentials (see Configuration) and start the server:

npm start

3. Connect your MCP client by adding a server entry that points to your local build, for example:

{
  "mcpServers": {
    "anysite-local": {
      "command": "node",
      "args": ["/path/to/anysite-mcp-server/build/index.js"],
      "env": {
        "ANYSITE_ACCESS_TOKEN": "your_token",
        "ANYSITE_ACCOUNT_ID": "your_account_id"
      }
    }
  }
}

Configuration examples and environment variables are described in the repository's README and docs.

Tips and common considerations:

• The remote MCP setup relies on OAuth; ensure ANYSITE_OAUTH_URL is correctly set to the MCP OAuth endpoint provided by your instance.
• If running locally, you may need ANYSITE_ACCESS_TOKEN and ANYSITE_ACCOUNT_ID (per your account settings) to access API features.
• The tool catalog includes 65+ tools; start with core capabilities like search_linkedin_users, get_linkedin_profile, get_instagram_user, search_reddit_posts, duckduckgo_search, search_youtube_videos, and parse_webpage, then expand as needed.
• For MCP clients (Cline, Cursor, Windsurf, Claude Desktop, etc.), place the mcp_config.json in the standard path described in the README for seamless discovery.
• If you encounter rate limits or platform changes, rely on the server’s self-healing and backoff features and consider enabling proxy rotation if high-volume requests are required.