Jina AI MCP server
Jina AI's official remote server gives agents web search, URL-to-markdown reading, reranking, and embeddings-powered tools.
Jina AI MCP is the official remote server from Jina AI, the company behind the popular Reader (r.jina.ai), Search, Embeddings, and Reranker APIs. It bundles that whole search foundation into one endpoint so an agent can search the open web, read any URL as clean markdown, capture screenshots, and pull academic papers from arXiv and SSRN — then refine results with embeddings-backed tools that rerank documents by relevance, classify text, and deduplicate near-identical strings or images. Parallel variants of the search and read tools fan out multiple queries or URLs at once for broad topic coverage in a single round-trip.
The server runs at https://mcp.jina.ai/v1 over Streamable HTTP and authenticates with a Jina API key passed as an Authorization Bearer header. Some tools work without a key, but a key raises rate limits and unlocks the full toolset. Clients that lack native remote MCP support can bridge to it with the mcp-remote proxy.
Quick install
Copy-paste configs are provided for all 8 supported clients. Pick your client below.
Add to ~/.claude.json
{
"mcpServers": {
"jina": {
"type": "http",
"url": "https://mcp.jina.ai/v1",
"headers": {
"Authorization": "Bearer <JINA_API_KEY>"
}
}
}
}claude mcp add --transport http jina https://mcp.jina.ai/v1Available tools
| Tool | Description |
|---|---|
| primer | Gets up-to-date contextual information for the session, such as the current time and locale, for time-aware responses. |
| read_url | Extracts and converts a web page's content to clean, readable markdown. |
| parallel_read_url | Reads multiple web pages in parallel to extract clean content efficiently. |
| capture_screenshot_url | Captures a high-quality screenshot of a web page as a base64-encoded JPEG. |
| guess_datetime_url | Estimates a web page's last-updated or published datetime from headers, metadata, and visible dates. |
| search_web | Searches the entire web for current information, news, articles, and websites. |
| parallel_search_web | Runs multiple web searches in parallel for comprehensive topic coverage. |
| search_images | Searches for images across the web, similar to Google Images. |
| search_arxiv | Searches academic papers and preprints on the arXiv repository. |
| parallel_search_arxiv | Runs multiple arXiv searches in parallel for broad research coverage. |
| search_ssrn | Searches papers on SSRN (Social Science Research Network) across social science, economics, law, and finance. |
| parallel_search_ssrn | Runs multiple SSRN searches in parallel for broad social-science research coverage. |
| search_bibtex | Searches DBLP and Semantic Scholar for academic papers and returns BibTeX citations. |
| search_jina_blog | Searches Jina AI news and blog posts for articles about AI, embeddings, neural search, and Jina products. |
| expand_query | Expands and rewrites a search query into multiple variants using a query-expansion model. |
| sort_by_relevance | Reranks a list of documents by relevance to a query using the Jina Reranker API. |
| classify_text | Classifies texts into user-defined labels using Jina embeddings. |
| deduplicate_strings | Returns the top-k semantically unique strings from a list using embeddings and submodular optimization. |
| deduplicate_images | Returns the top-k semantically unique images using Jina CLIP v2 embeddings and submodular optimization. |
| extract_pdf | Extracts figures, tables, and equations from PDF documents using layout detection. |
| show_api_key | Returns the bearer token from the Authorization header for debugging the MCP configuration. |
Required configuration
- JINA_API_KEYOptional
Jina AI API key (from jina.ai). Optional for some tools; raises rate limits and unlocks the full toolset.
What you can do with it
Research with reading built in
An agent searches the web, reads the most promising URLs as clean markdown, and reranks the passages by relevance before answering — all from one server.
Academic literature review
Fan out parallel arXiv and SSRN searches, pull BibTeX citations, and extract tables and equations from the source PDFs to assemble a grounded summary.
FAQ
- Is it free?
- The server is open source and several tools work without a key; a JINA_API_KEY raises rate limits and unlocks the full toolset for production use.
- Does it support remote/OAuth?
- It is a hosted remote endpoint at https://mcp.jina.ai/v1 over Streamable HTTP, authenticated with a Jina API key as an Authorization Bearer token rather than OAuth. There is no local stdio package; use the mcp-remote proxy for clients without native remote support.