Jina AI Remote

You run a remote MCP server that exposes Jina Reader, Embeddings, and Reranker APIs to your client applications. It enables a wide range of web, image, and document tools accessible through a streamlined, Streamable HTTP transport. This guide shows practical steps to use and install the server, plus important configuration and troubleshooting tips.

How to use

Connect your MCP-enabled client to the remote server by adding an MCP server entry that points to the remote endpoint. You have two connection options: a direct HTTP/remote configuration or a local proxy that talks to the remote MCP server via a local command.

If your client supports remote MCP servers, configure the server URL and optional authorization header as shown in the HTTP example. If your client does not yet support remote MCP servers, you can use a local proxy with a command that starts the proxy and forwards to the remote MCP endpoint.

Example HTTP connection (remote MCP URL). This uses a bearer token header for authorization when your client supports remote MCP servers.

Config examples for connecting to the MCP server

{
  "mcpServers": {
    "jina-mcp-server": {
      "url": "https://mcp.jina.ai/v1",
      "headers": {
        "Authorization": "Bearer ${JINA_API_KEY}" // optional
      }
    }
  }
}

Local proxy connection using mcp-remote

{
  "mcpServers": {
    "jina-mcp-server": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.jina.ai/v1"
        "--header",
        "Authorization: Bearer ${JINA_API_KEY}"
        ]
    }
  }
}

OpenAI Claude and Codex integration notes

If you are integrating Claude Code or Claude OpenAI Codex, follow the explicit commands shown for adding the MCP server or configuring the local proxy. Keep the transport and header formatting exact to avoid parsing issues in Windows environments.

Tool filtering before registering

You can filter which tools are registered with your MCP client by using query parameters on the MCP URL. This helps preserve your model’s context window by excluding tools that you do not need.

Key query parameters include exclude_tools, include_tools, exclude_tags, and include_tags. Tags categorize tools, such as read, search, parallel, and rerank, enabling precise control over what the client loads.

Troubleshooting

If you encounter a tool-calling loop, ensure your model has enough context length to hold the entire tool chain and reasoning. This is common with smaller models or those with limited context windows.

If not all tools appear to be available, refresh tool definitions in your MCP client. Some clients cache tool definitions locally and may require reloading or re-adding the MCP server to update the tool list.

For Windows users, a known issue can occur when spaces inside arguments are not escaped. Use the exact command format and pass the header via an environment variable when possible, to avoid parsing issues.

Developer and local development notes

To run and develop locally, you typically install dependencies and start the server in a development environment. The steps generally involve cloning the project, installing dependencies, and starting the server with the appropriate script.

Deployment to a hosting environment can be pursued via standard cloud deployment workflows. For example, you can deploy such servers to a cloud worker platform or run them behind a proxy that forwards requests to the MCP endpoint.

Available tools

primer

Get current contextual information for localized, time-aware responses

read_url

Extract clean, structured content from web pages as markdown via Reader API

capture_screenshot_url

Capture high-quality screenshots of web pages via Reader API

guess_datetime_url

Analyze web pages for last update/publish datetime with confidence scores

search_web

Search the entire web for current information and news via Reader API

search_arxiv

Search academic papers and preprints on arXiv via Reader API

search_ssrn

Search academic papers on SSRN via Reader API

search_images

Search for images across the web via Reader API

search_jina_blog

Search Jina AI news and blog posts at jina.ai/news

expand_query

Expand and rewrite search queries based on the query expansion model via Reader API

parallel_read_url

Read multiple web pages in parallel for efficient content extraction via Reader API

parallel_search_web

Run multiple web searches in parallel via Reader API

parallel_search_arxiv

Run multiple arXiv searches in parallel via Reader API

parallel_search_ssrn

Run multiple SSRN searches in parallel via Reader API

sort_by_relevance

Rerank documents by relevance via Reranker API

deduplicate_strings

Get top-k semantically unique strings via Embeddings API

deduplicate_images

Get top-k semantically unique images via Embeddings API

extract_pdf

Extract figures, tables, and equations from PDFs using layout detection