Skip to content
Gridlode
Docs · Reference

AI Data Extraction

GET /v1/ai/extractUsage-basedcache 60 min

Structured JSON from any web page, extracted by Claude. Give a URL and either a JSON Schema or a plain list of the fields you want (with optional instructions); the page is fetched (optionally rendered in Chrome), reduced to its main content as Markdown and read by the model, which only returns values that appear on the page. Billed by the model's actual token usage plus one credit for the page, never above your max_credits.

Query parameters

url
string · required
Public http(s) URL of the page to extract from.
schema
string
JSON Schema of the object to return. Give schema or fields.
Default: null
fields
string
Field names to extract, optionally with a description: ['company', 'price: monthly price of the cheapest plan'] or {'company': 'company name'}. Give schema or fields.
Default: null
instructions
string
Extra guidance for the extraction, e.g. 'prices in EUR only'.
Default: null
render
boolean
Render JavaScript in real Chrome before reading the page (+1 credit).
Default: false
wait_ms
integer
With render=true: extra wait after load.
Default: 1000
content
string
'main' = the page's main content as Markdown (boilerplate removed); 'full' = all visible text.
Default: "main"
Allowed: main, full
include_links
boolean
Keep link URLs in the text sent to the model.
Default: false
respect_robots
boolean
Refuse URLs disallowed by the site's robots.txt.
Default: true
max_chars
string
Truncate the page text to this many characters (the server limit applies when lower; default 60,000).
Default: null
max_credits
integer
The most this call may cost. The actual charge follows the model's token usage and is never higher.
Default: 150
max_age
integer
Accept a cached answer up to this many seconds old, billed at 50%.

Response data fields

    Example

    curl "https://api.gridlode.com/v1/ai/extract?" \
  -H "Authorization: Bearer $GRIDLODE_KEY"

    Try it without code on the AI Data Extraction API page.