Jina AI Reader & Search
OfficialOfficial Jina AI remote MCP server — read web pages as Markdown and run grounded web search over HTTP.
Add to your client
Copy the config for your MCP client and paste it into its config file.
Add remote MCP URL https://mcp.jina.ai/v1 with Authorization: Bearer <jina-api-key>Paste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"jina-ai-reader-search": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.jina.ai/v1",
"--header",
"Authorization: Bearer <your-jina-api-key>"
]
}
}
}Claude Desktop connects to remote servers through the `mcp-remote` proxy (installed on first run via npx). Restart Claude Desktop after saving.
Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- A Jina API key (free tier) from https://jina.ai — required for search and reranker tools; some read tools work key-less at lower rate limits
- An MCP client with remote/HTTP support (e.g. Claude Desktop, Cursor), or `mcp-remote` as a local proxy for clients without it
- Network access to the hosted endpoint `https://mcp.jina.ai/v1` — no local install needed
About Jina AI Reader & Search
This is Jina AI's official remote MCP server, exposing Jina's Reader, Search, Embeddings, and Reranker APIs as a single set of MCP tools over HTTP. It lets an agent read web pages as clean Markdown, run grounded web/image/academic search, and do downstream operations like reranking, classification, and deduplication — all without running anything locally.
It is a hosted Streamable HTTP server at https://mcp.jina.ai/v1 (an /sse endpoint is kept for backward compatibility). Clients with native remote-MCP support (Claude Desktop, Cursor) connect directly with a URL and a Bearer token; clients without native support bridge through the mcp-remote local proxy.
The server exposes around 19 tools spanning reading (read_url, parallel_read_url, capture_screenshot_url), search (search_web, search_images, search_arxiv, plus parallel variants), and AI processing (sort_by_relevance, classify_text, deduplicate_strings, expand_query, extract_pdf). Some reading tools work without a key at reduced rate limits; search and reranker tools require a free Jina API key.
Tools & capabilities (12)
read_urlExtract the content of a web page as clean Markdown (Jina Reader).
parallel_read_urlRead multiple URLs concurrently.
capture_screenshot_urlCapture a screenshot of a web page.
search_webRun a grounded web search and return results with content.
search_imagesSearch for images across the web.
search_arxivSearch academic papers and preprints on arXiv.
parallel_search_webRun multiple web searches concurrently.
sort_by_relevanceRerank a set of documents against a query using the Jina Reranker.
classify_textClassify text using Jina Embeddings.
deduplicate_stringsRemove semantically duplicate strings.
expand_queryExpand a query into related variants to improve retrieval.
extract_pdfExtract figures and tables from a PDF.
When to use it
- Use it when an agent needs to read web pages as Markdown without managing a local fetcher or browser.
- Use it when you want grounded web, image, or arXiv search results fed straight into a model.
- Use it when you need to rerank or deduplicate retrieved documents before passing them to an LLM (RAG pipelines).
- Use it when you want a fully hosted, no-local-install web-intelligence toolset over HTTP.
- Use it when reading or searching many URLs at once via the parallel tool variants.
Quick setup
- 1Sign up at jina.ai and copy your free Jina API key.
- 2Add the server URL `https://mcp.jina.ai/v1` to your MCP client config with an `Authorization: Bearer <JINA_API_KEY>` header.
- 3For clients without native remote-MCP support, wrap the connection with the `mcp-remote` proxy.
- 4Restart the client so it discovers the Jina tools.
- 5Verify by asking the model to read_url a known page or run a search_web query.
Security notes
Pass the JINA_API_KEY as a Bearer token stored in your client's secret config, not in shared URLs or logs. Queries and fetched pages pass through Jina's hosted infrastructure, so avoid sending sensitive data in requests.
Jina AI Reader & Search FAQ
Do I need an API key?
Search and reranker tools require a free Jina API key. Some reading tools (like read_url) work without one, but at reduced rate limits.
Where do I get the key and is it free?
Get it at https://jina.ai; there is a free tier. Higher throughput and the search tools require the key in an Authorization: Bearer header.
Is this remote or local?
It is a hosted remote server at https://mcp.jina.ai/v1 over Streamable HTTP. There is nothing to install locally.
My client doesn't support remote MCP — what now?
Use the mcp-remote proxy locally to bridge a stdio client to the hosted HTTP endpoint.
What's the difference between read_url and search_web?
read_url fetches and returns the Markdown of a specific page; search_web runs a web query and returns matching results with content.
Alternatives to Jina AI Reader & Search
Compare all alternatives →
Official MCP reference server that fetches a URL and returns its content as clean Markdown, with chunking.
Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.
Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.
Compare Jina AI Reader & Search with: