| name | jina-reader |
| description | Fetch a webpage as clean, structured Markdown via the jina.ai reader (r.jina.ai) instead of raw HTML — use whenever you need to read web content and want to save tokens, or when a normal fetch is blocked (HTTP 403/401), returns messy/boilerplate HTML, or the page is JavaScript-rendered (SPA, docs sites, wikis like grokipedia/notion/medium). Triggers: "fetch this page", "read this URL", "summarize this article/webpage", "get the docs at <url>", or any WebFetch that returned 403/garbled HTML. |
jina.ai reader — clean Markdown from any URL
The jina.ai reader proxies a target page and returns readable Markdown (article text, headings, tables, code — boilerplate/nav/ads stripped). This is far cheaper in tokens than raw HTML and often succeeds where a direct fetch is blocked or the page is JS-rendered.
How to use
Prefix the target URL with https://r.jina.ai/ and fetch that, then run your prompt against the result:
WebFetch(
url: "https://r.jina.ai/https://example.com/some/article",
prompt: "<what you want extracted>"
)
Rules:
- Keep the target URL's scheme in the path:
https://r.jina.ai/https://… (or http://…). Do not URL-encode it.
- Works for both
WebFetch and a raw curl -s "https://r.jina.ai/https://…" in Bash when you need the full text rather than an LLM extraction.
When to reach for it (vs a plain fetch)
- A direct
WebFetch returned HTTP 403/401 or a cross-host redirect loop.
- The page is a JS-heavy SPA / wiki / docs site (grokipedia, notion, medium, confluence, gitbook) where raw HTML is mostly script.
- You want structured Markdown to save tokens — long articles, API docs, spec pages, tables.
- You only need the readable content, not the exact HTML/DOM.
When NOT to use it
- Authenticated/private URLs (jina can't see your session) — use the proper authenticated tool (
gh, an MCP fetch, Google Drive/Gmail tools).
- You specifically need the raw bytes/HTML/JSON of an API or file (fetch the URL directly).
- Very large pages you only need a slice of — fetch directly with a narrow prompt if the direct fetch works.
Caveats
- It's a third-party proxy: the target URL (but not your auth) is sent to jina.ai. Don't route sensitive/internal URLs through it.
- Content is a rendered snapshot; for fast-changing pages re-fetch rather than trusting a cache.
- If jina itself is rate-limited or errors, fall back to a direct
WebFetch/curl, or WebSearch.
