# Jina

> Fetch web pages or search the web via Jina AI (r.jina.ai / s.jina.ai). Use when WebFetch fails, returns empty/garbage, or for JS-heavy sites: x.com, Notion, SPAs, PDFs at a URL, Cloudflare-gated pages. Also use for grounded web search with full-page content. Triggers on "fetch this URL", "read this page", "search for", or after a WebFetch error. Do NOT use for simple static HTML.

- Skill: `chrisliu298/jina` (Agent Skill)
- Install (CLI): `npx skillmds@latest add chrisliu298/jina`
- Raw SKILL.md: https://api.skillmd.com/api/skills/chrisliu298/jina/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Web & Frontend
- Author: chrisliu298 (https://skillmd.com/u/chrisliu298)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/chrisliu298/jina

---


# Jina

Fetch web content and search the web using Jina AI's Reader and Search APIs.

## When to use

- WebFetch returned an error, empty content, or garbled HTML
- The URL points to a JS-rendered page (x.com, Notion, SPAs, React/Vue apps)
- The URL points to a PDF
- You need web search results with full page content (not just titles/snippets)

## Reader — fetch a URL

```bash
curl -s -X POST 'https://r.jina.ai/' \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json' \
  -H 'X-Timeout: 30' \
  ${JINA_API_KEY:+-H "Authorization: Bearer $JINA_API_KEY"} \
  -d '{"url": "TARGET_URL"}'
```

Always use POST (not the GET prefix `r.jina.ai/URL`) because GET silently drops `#` fragments, breaking SPA URLs.

### Response

JSON with `data.content` (markdown), `data.title`, `data.url`, `data.usage.tokens`.
Check `data.warning` — Reader returns HTTP 200 even when the target site errors (e.g., 404). A non-empty `warning` field means something went wrong.

### Useful headers

Add these as `-H 'Header: value'` to the curl command when needed:

| Header | When to use |
|--------|------------|
| `X-Wait-For-Selector: <css>` | SPA content loads after JS — wait for a specific element |
| `X-Target-Selector: <css>` | Extract only a specific part of the page |
| `X-Remove-Selector: <css>` | Strip navbars, ads, cookie banners before extraction |
| `X-No-Cache: true` | Need fresh content (default cache is ~1 hour) |
| `X-With-Links-Summary: true` | Append a list of all links found on the page |
| `X-With-Images-Summary: all` | Append a list of all image URLs (useful since images are lost in markdown) |
| `X-Retain-Images: none` | Strip all images to reduce output size |
| `X-Engine: direct` | Skip browser rendering (faster, but misses JS content) |

For **x.com/Twitter** specifically, add:
```
-H 'X-Wait-For-Selector: [data-testid="tweetText"]' \
-H 'X-With-Images-Summary: all'
```

## Search — web search with full content

```bash
curl -s 'https://s.jina.ai/QUERY_URL_ENCODED' \
  -H 'Accept: application/json' \
  -H "Authorization: Bearer $JINA_API_KEY"
```

An API key is **required** for search (free key works — get one at https://jina.ai/?sui=apikey).

Returns top 5 results, each with `title`, `content` (full page markdown), and `url`.

To restrict to specific sites, append `?site=example.com`.

## Parsing the response

Pipe through python to extract the content:

```bash
# Reader — extract markdown content
... | python3 -c "
import sys, json
r = json.load(sys.stdin)
d = r.get('data', {})
if d.get('warning'):
    print('WARNING:', d['warning'], file=sys.stderr)
print(d.get('content', d.get('text', '')))"

# Search — extract all results
... | python3 -c "
import sys, json
r = json.load(sys.stdin)
for i, item in enumerate(r.get('data', []), 1):
    print(f'## Result {i}: {item.get(\"title\", \"\")}')
    print(f'URL: {item.get(\"url\", \"\")}')
    print(item.get('content', ''))
    print()"
```

## Limitations

- **Images are lost** — pages convert to markdown text. Image URLs may appear via `X-With-Images-Summary` but the images themselves are not rendered.
- **Login-required pages fail** — Reader cannot authenticate. Paying more does NOT unlock more sites.
- **Bot-protected sites may fail** — Cloudflare challenges, aggressive CAPTCHAs. Reader does not bypass anti-bot measures.
- **Search costs 10K tokens minimum** per request regardless of output size.
- **Rate limits without key**: Reader ~20 RPM, Search blocked. With free key: Reader 500 RPM, Search 100 RPM.