Jina
Any URL as clean, LLM-ready markdown (Reader), and web search that returns each result's page text.
- Provider id:
jina - Tools: 2, at $0.005–$0.01 per call
- Categories: Content Extraction, Web Search
- Homepage: jina.ai/reader
- Provider docs: r.jina.ai/docs
| Tool | Price | What it does |
|---|---|---|
jina/read | $0.005 | Turn any URL (HTML page or PDF) into clean markdown with its title, language and publish date. |
jina/search | $0.01 | Web search that returns the top results with each page's text as markdown, in one call. |
Inputs are JSON bodies, and unknown fields are rejected with 422 invalid_input (not charged). POST /v1/inspect with a tool's id returns the same input schema, plus the output schema.
Jina Reader
jina/read · $0.005 per call (standard) · Content Extraction
Turn any URL (HTML page or PDF) into clean markdown with its title, language and publish date.
Fetches one URL and returns its main content as markdown without navigation, ads or images, plus title, description, language, published time and the HTTP status the site answered (404 or 410 comes back as found: false). Fast by default because it does not run JavaScript; set render_js for single-page apps. One URL per call and no search: find pages with serper/search, or get several results' text at once with jina/search. For a page that blocks plain fetchers, firecrawl/scrape renders it in a full browser.
Related: jina/search, firecrawl/scrape, groq/summarize, serper/search.
Input
| Field | Type | Required | Description |
|---|---|---|---|
url | string (uri) | yes | The page or PDF to read. 1–2083 characters. |
render_js | boolean | no | Load the page in a headless browser first. Slower (often 5–15 s, may hit the deadline); only for pages that show nothing without JavaScript. Default false. |
keep_links | boolean | no | Keep hyperlinks as markdown links (default: link text only). Default false. |
target_selector | string | no | CSS selector: return only this part, e.g. 'article' or '#main'. At most 200 characters. |
max_tokens | integer | no | Cut the page after about this many tokens (4 characters each). From 500 to 5000. Default 5000. |
timeout_s | integer | no | Give up on a slow site after this many seconds (default: as long as the run deadline allows). From 1 to 7. |
Example
{
"url": "https://en.wikipedia.org/wiki/Transistor",
"max_tokens": 2000
}node akashi.mjs run jina/read --input '{"url":"https://en.wikipedia.org/wiki/Transistor","max_tokens":2000}'Jina Search
jina/search · $0.01 per call (premium) · Web Search, Content Extraction
Web search that returns the top results with each page's text as markdown, in one call.
Searches the web and reads every result for you: each result has its title, url, description and up to ~4,000 characters of the page's main text. Use it when you need to read several sources, not just pick one. Slower and dearer than plain search (Jina bills 10k tokens per query), so for links and snippets only use serper/search; for one known URL use jina/read; for a written, cited answer use akashi/answer.
Related: serper/search, jina/read, akashi/answer.
Input
| Field | Type | Required | Description |
|---|---|---|---|
query | string | yes | What to search for. 1–400 characters. |
num | integer | no | Results to return, each with text. From 1 to 10. Default 5. |
site | string | no | Only results from this domain, e.g. 'docs.python.org'. At most 253 characters. Pattern ^[A-Za-z0-9.-]+\.[A-Za-z]{2,}$. |
country | string | no | Two-letter country, e.g. 'us'. Pattern ^[A-Za-z]{2}$. |
language | string | no | Result language, e.g. 'en'. Pattern ^[A-Za-z]{2,3}(-[A-Za-z]{2,4})?$. |
Example
{
"query": "how do solid-state batteries work",
"num": 3
}node akashi.mjs run jina/search --input '{"query":"how do solid-state batteries work","num":3}'