mcp-jinaai-reader

This MCP server integrates Jina.ai's Reader API to extract and convert web page content into a format optimized for large language models. The server exposes a single MCP tool named read_url, which takes a URL and returns structured, LLM-friendly text with options to control formatting, caching, and content scope. To use it, configure your MCP client to point at the jinaai-reader server and supply your Jina.ai API key. Once running, you can request content from any URL, with the response tailored for downstream NLP tasks such as summarization, indexing, or document understanding. The tool supports extracting and formatting documentation, web pages, and other content types while preserving structure and enabling optional summaries of links or images.

Prerequisites:

• Node.js (recommended LTS) and npm installed
• Access to a Jina.ai API key

Installation steps:

1. Install dependencies and build the project

# Clone the repository (adjust path if needed)
git clone https://github.com/spences10/mcp-jinaai-reader.git
cd mcp-jinaai-reader

# Install dependencies
npm install

2. Build the project (if required by the repo)

npm run build

3. Run the server in development mode

npm run dev

4. Run in production (example)

npm run build
npm start

Notes:

• Ensure you have a valid Jina.ai API key and set it in the environment variable JINAAI_API_KEY when running the server or in your MCP client configuration.
• The README examples show how to configure MCP clients for different environments; adapt commands to your deployment environment as needed.

Tips and considerations:

• The JINAAI_API_KEY environment variable is required for authentication with the Jina.ai Reader API.
• In non-Windows environments, you may run the node server directly or via an MCP orchestrator; the tool configuration uses the read_url command with URL, and optional parameters for formatting, timing, and selective content extraction.
• If you encounter authentication errors, double-check that the API key is correctly exported in the environment where the MCP client launches the server.
• The server supports options like no_cache, format (json or stream), timeout, and various CSS selector-based filters to focus or exclude content.
• When integrating with Claude Desktop or WSL, mirror the config example provided in the README to ensure proper key propagation.