Firecrawl: turn a page into the exact JSON you asked for

The Stack

Tested Against

Prerequisites

• A Firecrawl API key or a self-hosted instance (only to run the crawl)

Steps

1. 1

   Author the extract request and validate it

   Write the extract request: the url, formats including json, and a jsonOptions.schema describing the fields you want. CI parses it and checks the shape. Sending it to Firecrawl needs a key or your self-hosted instance, so that step is fenced.

   cat > extract.json <<'JSON'
   {
     "url": "https://example.com/pricing",
     "formats": ["markdown", "json"],
     "onlyMainContent": true,
     "jsonOptions": {
       "schema": {
         "type": "object",
         "properties": {
           "plan": { "type": "string" },
           "priceUsd": { "type": "number" }
         },
         "required": ["plan", "priceUsd"]
       }
     }
   }
   JSON
   node -e '
   const fs = require("fs");
   const c = JSON.parse(fs.readFileSync("extract.json", "utf8"));
   function bad(m) { console.error("BAD: " + m); process.exit(1); }
   if (!c.url || typeof c.url !== "string") bad("request has no url");
   if (!Array.isArray(c.formats) || !c.formats.includes("json")) bad("formats must include json for structured extraction");
   const sch = c.jsonOptions && c.jsonOptions.schema;
   if (!sch || sch.type !== "object" || !sch.properties) bad("no jsonOptions.schema object with properties");
   const fields = Object.keys(sch.properties).length;
   if (fields < 1) bad("schema defines no fields");
   console.log("config OK: Firecrawl extract targets a url, requests json output, and defines a " + fields + "-field schema");
   '

2. 2

   Run the crawl (the network step, not checked by CI)

   Send the request to the Firecrawl API with your key, or to your self-hosted instance. Firecrawl renders the page and returns JSON matching your schema. Only crawl pages you are allowed to; the network call and its data are fenced.