mcp-jinaai-search

View original on GitHub 

A Model Context Protocol (MCP) server for integrating Jina.ai’s Search API with LLMs. This server provides efficient and comprehensive web search capabilities, optimised for retrieving clean, LLM-friendly content from the web.

Features

• 🔍 Advanced web search through Jina.ai Search API
• 🚀 Fast and efficient content retrieval
• 📄 Clean text extraction with preserved structure
• 🧠 Content optimised for LLMs
• 🌐 Support for various content types including documentation
• 🏗️ Built on the Model Context Protocol
• 🔄 Configurable caching for performance
• 🖼️ Optional image and link gathering
• 🌍 Localisation support through browser locale
• 🎯 Token budget control for response size

Configuration

This server requires configuration through your MCP client. Here are examples for different environments:

Cline Configuration

Add this to your Cline MCP settings:

{
	"mcpServers": {
		"jinaai-search": {
			"command": "node",
			"args": ["-y", "mcp-jinaai-search"],
			"env": {
				"JINAAI_API_KEY": "your-jinaai-api-key"
			}
		}
	}
}

Claude Desktop with WSL Configuration

For WSL environments, add this to your Claude Desktop configuration:

{
	"mcpServers": {
		"jinaai-search": {
			"command": "wsl.exe",
			"args": [
				"bash",
				"-c",
				"JINAAI_API_KEY=your-jinaai-api-key npx mcp-jinaai-search"
			]
		}
	}
}

Environment Variables

The server requires the following environment variable:

• JINAAI_API_KEY: Your Jina.ai API key (required)

API

The server implements a single MCP tool with configurable parameters:

search

Search the web and get clean, LLM-friendly content using Jina.ai Reader. Returns top 5 results with URLs and clean content.

Parameters:

• query (string, required): Search query
• format (string, optional): Response format (“json” or “text”). Defaults to “text”
• no_cache (boolean, optional): Bypass cache for fresh results. Defaults to false
• token_budget (number, optional): Maximum number of tokens for this request
• browser_locale (string, optional): Browser locale for rendering content
• stream (boolean, optional): Enable stream mode for large pages. Defaults to false
• gather_links (boolean, optional): Gather all links at the end of response. Defaults to false
• gather_images (boolean, optional): Gather all images at the end of response. Defaults to false
• image_caption (boolean, optional): Caption images in the content. Defaults to false
• enable_iframe (boolean, optional): Extract content from iframes. Defaults to false
• enable_shadow_dom (boolean, optional): Extract content from shadow DOM. Defaults to false
• resolve_redirects (boolean, optional): Follow redirect chains to final URL. Defaults to true

Development

Setup

1. Clone the repository
2. Install dependencies:

pnpm install

3. Build the project:

pnpm run build

4. Run in development mode:

pnpm run dev

Publishing

1. Create a changeset:

pnpm changeset

2. Version the package:

pnpm version

3. Build and publish:

pnpm release

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - see the LICENSE file for details.

Acknowledgments

• Built on the Model Context Protocol 
• Powered by Jina.ai Search API