MCP

This MCP server provides access to Jina Reader, Embeddings, and Reranker APIs via a remote MCP endpoint. It exposes a wide suite of tools that can perform web content extraction, web searches, academic paper lookups, image searches, and reranking/deduplication tasks. Clients can connect using the MCP transport (Streamable HTTP) and then call the registered tools such as read_url, search_web, search_images, deduplicate_strings, and sort_by_relevance, among others. If you’re using the remote MCP configuration, you’ll route requests through the mcp-remote proxy to the official remote endpoint at mcp.jina.ai/v1 and pass your API key when needed.

Prerequisites:

• Node.js and npm installed on your machine
• Internet access to install npm packages

Install steps:

1. Ensure you have an API key from Jina (optional for higher rate limits):
   • Obtain a free Jina API key from https://jina.ai
2. Install or run the MCP server via npm/npx using the remote proxy configuration provided by this server:
   • This configuration expects you to run the remote MCP via npx mcp-remote with the remote URL and your API key scope:

# Example to start the remote MCP proxy locally (using the provided config):
npx -y mcp-remote https://mcp.jina.ai/v1 --header "Authorization: Bearer ${JINA_API_KEY}"

3. Alternatively, if you want to run a local proxy pointing at the remote MCP endpoint, ensure you replace placeholders with your actual API key:

export JINA_API_KEY=your_api_key_here
npx -y mcp-remote https://mcp.jina.ai/v1 --header "Authorization: Bearer ${JINA_API_KEY}"

Notes:

• The server uses the Streamable HTTP transport as per MCP spec 2025-03-26.
• Some clients may not support environment variables, so consider hardcoding the API key in a secure manner if needed by your environment.

Tips and caveats:

• Optional tools require an API key for higher rate limits; you can run with no key but expect stricter quotas.
• If you encounter issues with env variables in certain clients, replace ${JINA_API_KEY} with your actual key in logs/requests.
• The MCP tooling supports filtering tools to optimize token usage; consider excluding parallel or heavy tools if you’re operating under tight token budgets.
• Ensure network access to https://mcp.jina.ai/v1 when using the remote proxy; corporate proxies may require additional configuration.
• For OpenAI Codex or Claude Code integrations, follow the specific configuration steps shown in the README to register the server appropriately.