getJSON reads any content a user or an agent points at, a web page or a file, and returns JSON in your schema, validated. Try and use it without an account.
getJSON reads a link or a file and returns it as JSON, the envelope. Add a JSON Schema and it extracts the fields you asked for with a model. Sources: X, Bluesky, Mastodon, Hacker News, GitHub, public Google Docs and Sheets, PDF, PNG, JPEG, WebP, plain text, CSV and any web page.
Developers who build or run agents, and the agents themselves. An agent meets getJSON when it hits a link or a file it cannot read.
Each result is validated against your schema before it comes back. A result that breaks it gets one retry, then the call fails with schema_mismatch and the fields that failed. A link that blocks the read comes back with a wall that says why.
In credits. Reading a link or a file costs 0 credits. An extraction costs 1 to 3 credits: a fixed mode has one fixed price, and auto picks the model by the document and costs at most 3. Every account with credits can send an inline schema with a call. Saved schemas, used by schema_id, need the Pro plan. A new account gets 5 free credits.
An AI agent can use getJSON without a human: it reads links with no key, gets a trial key with no email when it needs more, and hands a claim link to its user. The plain-text guide for agents is llms.txt. The API reference is at /docs, and keys, quickstart and the sandbox are on the developer portal.