Jina Reader

Fetch the markdown content of any webpage using Jina AI's Reader API (r.jina.ai), or search the web using Jina AI's Search API (s.jina.ai). Use this skill whenever you need to read a URL and get clean LLM-friendly markdown output, extract article/documentation content from a page, or search the web and retrieve rich results with content. Triggers on: read this URL, fetch this page, get the content of, what does this page say, summarize this link, search the web for, look up, or any request requiring clean markdown from a URL.

Jina Reader & Search Skill

Two APIs, zero API key required (optional key for higher rate limits).

Reader — Fetch a URL as Markdown

Simple (shell):

bash

curl -s "https://r.jina.ai/<URL>"

With script:

bash

python3 skills/jina-reader/scripts/jina_read.py <URL>
python3 skills/jina-reader/scripts/jina_read.py <URL> --no-images   # strip images
python3 skills/jina-reader/scripts/jina_read.py <URL> --json        # structured JSON response

With API key (set in .env):

bash

JINA_API_KEY=your_key python3 skills/jina-reader/scripts/jina_read.py <URL>

Useful request headers (for direct curl):

Header	Purpose
Authorization: Bearer <key>	Higher rate limits
X-Return-Format: markdown	Force markdown output
X-Remove-Selector: img	Strip images
Accept: application/json	JSON response with title, url, content

Search — Web Search via Jina

Simple (shell):

bash

curl -s "https://s.jina.ai/<URL-encoded-query>"

With script:

bash

python3 skills/jina-reader/scripts/jina_search.py "your query"
python3 skills/jina-reader/scripts/jina_search.py "your query" --json
python3 skills/jina-reader/scripts/jina_search.py "your query" --json --results 3

API Key

API	Without Key	With Key
Reader (r.jina.ai)	✅ Works (rate-limited)	Higher limits
Search (s.jina.ai)	❌ Requires key	✅ Works

Store as JINA_API_KEY in .env. Get a free key at: https://jina.ai/api-dashboard

No key for search? Use the duckduckgo-search skill as a fallback.

Notes

• Reader works on most pages including JS-heavy sites
• Search returns up to 5 results by default, each with title, URL, and content summary
• For very long pages, use --no-images to reduce token usage