Skip to content
Jina AI MCP Server

Jina AI MCP Server

Added to Onei AI
Apache-2.0

Summary

Jina AI's official remote MCP server: read pages and PDFs as Markdown, search the web, arXiv and SSRN, take screenshots, rerank and deduplicate.

Features

  • Read any URL or PDF as Markdown, or just the passages that answer a question
  • Web, image, arXiv and SSRN search
  • Reranking and semantic deduplication via Jina's models
  • Several tools work without an API key
  • Filter tools server-side to save context

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it, filling in the headers with your own values.
  3. Restart the agent — the "Jina AI MCP Server" tools will be available.
{
  "mcpServers": {
    "jina-ai": {
      "type": "http",
      "url": "https://mcp.jina.ai/v1",
      "headers": {
        "Authorization": "Bearer ${JINA_API_KEY}"
      }
    }
  }
}

Description

Jina AI's remote MCP server puts the company's Reader, Search, Embeddings and Reranker APIs behind one endpoint, https://mcp.jina.ai/v1. Its 12 tools turn any web page or PDF into clean Markdown (optionally returning only the passages that answer a question), capture page screenshots, guess when a page was published or last updated, search the web, images, arXiv and SSRN, rerank documents by relevance to a query, return the top-k semantically distinct strings from a list, and extract figures, tables and equations from PDFs.

Some tools work without a key: primer, guess_datetime_url and the blog search are free, and read_url and screenshots run keyless at lower rate limits. Search, reranking, deduplication and PDF extraction need a Jina API key, and free keys are available at jina.ai. Query parameters on the endpoint (include_tools, exclude_tags, max_tokens and others) drop tools before they are registered, so the client spends no context on tools it will not use.

Clients with remote MCP support connect to the URL directly, with the key as a bearer header; others go through mcp-remote. The server's source is on GitHub under Apache-2.0.

Official SerpApi MCP server: real-time Google, Bing, YouTube and other search results as structured JSON or token-saving markdown.

MCP: wigolo

by KnockOutEZ

Local-first web MCP server for AI agents — search, fetch, crawl, extract and deep research with no API keys, no cloud round-trip and no per-query bill.

MCP: Crawlbase MCP

by Crawlbase

Official Crawlbase MCP server giving agents live web access — fetch any URL as raw HTML, clean Markdown or a screenshot, with JS rendering and proxy rotation handled for you.

MCP: Scry

by Scry

Remote MCP server that lets an agent run read-only SQL over roughly 164 billion indexed documents from Reddit, Hacker News, arXiv, Wikipedia, Stack Exchange and prediction markets.

Browse all MCP servers →