Do I need to define a schema up front?

No — describe the fields you want in plain language in the instructions. Add a JSON Schema in settings.extract_schema_json only when you need an exact, fixed shape.

How is AI extraction priced?

A flat 5 points per call, regardless of page size — see Action costs on the Pricing page.

Can I use AI extraction inside an autonomous agent?

Yes — expose it as a tool your agent can call on any URL it encounters mid-task, and get back structured JSON instead of raw HTML.

AI extraction

Turn any page or document into clean, structured JSON.

Point AI extraction at a URL or document and get back clean, structured JSON — no brittle selectors, no HTML parsing, no scrapers to rewrite when a layout changes.

Get started View documentation

Extraction run

ext_7bK1p

Fetched & rendered page

example.com/company/acme

Stripped boilerplate

ads, nav, and scripts removed

Extracted fields

3 fields mapped

{
  "company_name": "Acme",
  "founded": "2019",
  "employees": "50-100"
}

Capabilities

What AI extraction handles for you

Handles rendered pages, not just static HTML — JavaScript-heavy sites included

Ads, navigation, and boilerplate stripped automatically before extraction

Consistent field mapping across pages with different layouts

Output ready for RAG pipelines, agent tool calls, or direct storage

How it works

Send a URL, get structured JSON back

Every AI extraction request runs the same four-step pipeline under the hood.

Send a URL

Point AI extraction at any public page or PDF, with instructions for the fields you want.

Page is rendered

JavaScript-heavy pages are rendered just like a browser would show them.

Content is cleaned and structured

Ads and boilerplate are removed, and the remaining content is mapped to fields.

Get structured JSON

Receive the extracted fields, ready to use or index.

Structured output

Every page comes back as structured JSON

AI extraction always returns typed, structured JSON — never raw HTML — so your pipeline can rely on a consistent shape no matter which page it points at.

Fields you describe in instructions, optionally pinned with a JSON Schema
No HTML, scripts, or boilerplate leaking into the output
Ready to drop into a database, RAG index, or agent tool call

Response

{
  "session_id": "11111111-1111-1111-1111-111111111111",
  "kind": "extract",
  "url": "https://example.com/company/acme",
  "machine": {
    "company_name": "Acme",
    "founded": "2019",
    "employees": "50-100"
  }
}

API or MCP, every framework

Use it however you build.

Illustrative example — exact endpoint paths and response schema are defined in the full API documentation.

curl https://api.neuralverge.ai/functions/v1/run-extract \
  -H "Authorization: Bearer <API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/company/acme",
    "instructions": "Extract the company name, founding year and employee range.",
    "settings": { "country_code": "us" }
  }'

FAQ

Questions, answered

Any public URL, including JavaScript-rendered pages — the page is rendered before extraction, so you're not limited to static HTML.

AI Extract costs 5 points per call — see Action costs on the Pricing page.

Enrichment sources →Pricing →

Start building with AI extraction

Get an API key and start making calls in minutes.

Get started