name: Jina AI Official MCP Server
source_code: https://github.com/jina-ai/MCP
description: Official Model Context Protocol server for Jina AI APIs
version: 1.11.1
package_name: jina-mcp
usage: |
  
  {
  	"mcpServers": {
  	"jina-mcp-server": {
  		"url": "https://mcp.jina.ai/v1",
  		"headers": {
  		"Authorization": "Bearer ${JINA_API_KEY}" // optional
  		}
  	}
  	}
  }
get_api_key: https://jina.ai/api-dashboard/
endpoints:
  v1: /v1 - Primary endpoint
  sse: /sse - Alias for /v1 (backward compatibility)
tool_filtering:
  description: Reduce token usage by filtering tools via query parameters
  parameters:
    exclude_tools: Comma-separated tool names to exclude (e.g., search_web,search_arxiv)
    include_tools: Comma-separated tool names to include
    exclude_tags: Comma-separated tags to exclude (e.g., search,read)
    include_tags: Comma-separated tags to include
    max_tokens: Cap the size of read_url responses in tokens (0 disables truncation)
  tags:
    search:
      - search_web
      - search_arxiv
      - search_ssrn
      - search_images
      - search_jina_blog
    read:
      - read_url
      - capture_screenshot_url
    utility:
      - primer
      - guess_datetime_url
      - extract_pdf
    rerank:
      - sort_by_relevance
      - deduplicate_strings
  examples:
    - /v1?exclude_tags=search - Exclude all search tools
    - /v1?include_tags=search,read - Only include search and read tools
    - /v1?exclude_tools=search_ssrn,search_images - Exclude specific tools
  precedence: exclude_tools > exclude_tags > include_tools > include_tags
api_key:
  not_needed: primer, read_url, capture_screenshot_url, guess_datetime_url,
    search_jina_blog
  needed: search_web, search_arxiv, search_ssrn, search_images, sort_by_relevance,
    deduplicate_strings, extract_pdf, and read_url with a question (without a
    key it returns the full page)
time_limits:
  search: 30s per query, up to 5 queries per call
  read_url: 30s per page, up to 5 pages per call
  read_url_with_question: 25s for the passages, then 30s for the full page if none came back
  capture_screenshot_url: 60s
  sort_by_relevance: 30s
  deduplicate_strings: 30s
  extract_pdf: 90s
errors:
  reporting: "A failed call has isError: true and says why in plain text. A call
    with several queries or URLs is an error only if all of them failed;
    otherwise each failure is reported next to the results that did come back."
  key_rejected: The message carries the reason the API gave, e.g. an unknown key
    or a key that billing refused.
  busy: When a service is rate limited or out of capacity the message says so,
    says whether it is about your key, and gives the wait the service asked for
    if it named one.
  read_url_fallback: "With a question, read_url returns passages (snippet_source:
    content). If none could be ranked it returns the whole page instead, marked
    snippet_source: full_content with a note."
tools:
  - primer - Current time, user timezone, location and client context
  - read_url - Read a web page or PDF as markdown; pass question for passages
    instead of the full body
  - capture_screenshot_url - Capture a screenshot of a web page
  - guess_datetime_url - Guess a page publish or last-update datetime with a
    confidence score
  - search_web - Search the web; returns titles, URLs and engine snippets; query
    accepts an array
  - search_arxiv - Search arXiv preprints; query accepts an array
  - search_ssrn - Search SSRN working papers; query accepts an array
  - search_images - Search the web for images
  - search_jina_blog - Search Jina AI news at jina.ai/news
  - sort_by_relevance - Rerank documents by relevance to a query
  - deduplicate_strings - Select top-k semantically distinct strings
  - extract_pdf - Extract figures, tables and equations from a PDF
