jina -tools

This MCP server exposes tools that integrate Jina AI APIs for web content access and search. The server can run in two transport modes: stdio (suitable for local integrations spawned by another application) and http (for remote deployments). By default, the stdio configuration enables the jina-reader (jina_reader) to extract content from web pages and the web search capabilities via the jina_search tool (standard endpoint). When using the HTTP transport, you can connect MCP clients to http://localhost:3000/mcp and pair with clients such as MCP Inspector, Claude Code, or VS Code. The available tools rotate based on the --search-endpoint setting: standard uses s.jina.ai, while vip uses svip.jina.ai, enabling VIP search if configured with an API key. The CLI also supports pagination through tokens-per-page and page-level content handling via the reader.

Prerequisites:

• Node.js v16 or higher
• Optional: Jina AI API key for enhanced features

Installation and run (stdio transport):

1. Ensure Node.js is installed and available in your PATH. You can verify with: node -v npm -v
2. Run the MCP server using npx (no local install required): npx jina-mcp-tools --transport stdio --tokens-per-page 15000 --search-endpoint standard
3. If you want to configure an API key for enhanced features, set the environment variable when launching the command: JINA_API_KEY=your_api_key_here npx jina-mcp-tools --transport stdio --tokens-per-page 15000 --search-endpoint standard

Running via HTTP transport (remote deployment):

1. Start the server with HTTP transport and a port (default 3000): JINA_API_KEY=your_api_key npx jina-mcp-tools --transport http --port 3000 or, without an API key (reader only): npx jina-mcp-tools --transport http --port 3000
2. Connect MCP clients to the server at http://localhost:3000/mcp.

If you prefer a pre-recorded configuration, you can create an mcp_config.json snippet like the following (stdio transport):

{
  "mcpServers": {
    "jina-mcp-tools": {
      "command": "npx",
      "args": [
        "jina-mcp-tools",
        "--transport", "stdio",
        "--tokens-per-page", "15000",
        "--search-endpoint", "standard"
      ],
      "env": {
        "JINA_API_KEY": "your_jina_api_key_here_optional"
      }
    }
  }
}

Tips and considerations:

• The Jina API key is optional for basic usage (reader-only). Enabling API-based features requires the key for higher feature access.
• The --search-endpoint option selects between standard (s.jina.ai) and vip (svip.jina.ai). If you need VIP search capabilities, provide a valid API key and use --search-endpoint vip.
• Tokens-per-page controls how many tokens are returned in a page for the reader; higher values can improve pagination but may impact latency.
• When running HTTP transport, ensure port 3000 is open and accessible to MCP clients. Use proper firewall rules if deployed behind NAT or a reverse proxy.
• The jina_reader tool automatically handles pagination for large documents and caches recent URLs for speed.
• If you need to modify the server configuration frequently, consider maintaining an mcp_config.json and referencing it in your startup script.