AI Web Extraction API

API ID 14423

Extract structured data from any web page with a plain-language instruction. Returns JSON plus a validation block that checks numeric values against the page text and their history.

100% uptime 2,435 ms avg response

API Documentation

Endpoints

Request
Fetches the page at the given URL, keeps it in cache for the window set by expiration_seconds, and asks a language model for the fields described in parse_description. Numeric answers are checked against the page text and against the previous value for the same URL and instruction; the check is returned in the validation object and never blocks the answer.
Endpoint ID: 36427
GET https://zylalabs.com/api/14423/ai+web+extraction+api/36427/extract+from+page
INPUT PARAMETERS

Extract From Page — Endpoint Features

Object Description
url Required Page to extract from: full URL including http:// or https://.
parse_description Required What to extract from the page, in plain language. Name the fields you want back.
expiration_seconds Optional Reuse a previous scrape and answer for the same URL and instruction if it is at most this many seconds old. Default 600 (10 minutes). 0 forces a fresh scrape.
raw_text Optional true: the model also returns the value exactly as written on the page, in data and in validation.raw_text. Default false.
value_field Optional Key of the answer that holds the number to check, when the answer has more than one number. Default: the answer's single number.
last_value Optional Your last known value for this URL and instruction; big jumps against it are flagged in validation. Default: the service's own history.
max_chars Optional Characters of page text sent to the model after pruning. Default 6000. 0 sends the whole page.

Free test requests remaining: 3 of 3.


INPUT PARAMETERS

url
parse_description
expiration_seconds
raw_text
value_field
last_value
max_chars
API EXAMPLE RESPONSE
JSON
{
    "data": [
        {
            "value": "4010.00",
            "raw_text": "4,010.00"
        }
    ],
    "validation": {
        "status": "ok",
        "reason": "no_reference",
        "value": 4010,
        "model_value": 4010,
        "raw_text": "4,010.00",
        "reference": null,
        "change_pct": null,
        "suspect_pct": 10,
        "reject_pct": 30,
        "on_page": true,
        "votes": [],
        "mode": "shadow",
        "reference_source": null
    }
}
Extract From Page — CODE SNIPPETS

curl --location --request GET 'https://zylalabs.com/api/14423/ai+web+extraction+api/36427/extract+from+page?url=https://www.investing.com/commodities/gold&parse_description=Current gold futures price as a plain number, in a field called value&expiration_seconds=600&raw_text=true&value_field=value&last_value=4012.5&max_chars=6000' --header 'Authorization: Bearer YOUR_API_KEY' 


    
Request
Renders the page, takes a full-page screenshot and asks a vision model for the fields described in llm_instructions. Use it for pages whose data only appears once rendered: charts, widgets, canvas tables or heavy JavaScript apps. Numeric answers are checked against their history and reported in the validation object.
Endpoint ID: 36428
GET https://zylalabs.com/api/14423/ai+web+extraction+api/36428/extract+from+screenshot
INPUT PARAMETERS

Extract From Screenshot — Endpoint Features

Object Description
url Required Page to extract from: full URL including http:// or https://.
llm_instructions Required What to read from a screenshot of the page, in plain language. Name the fields you want back.
expiration_seconds Optional Reuse a previous screenshot and answer for the same URL and instruction if it is at most this many seconds old. Default 600 (10 minutes). 0 forces a fresh capture.
value_field Optional Key of the answer that holds the number to check, when the answer has more than one number. Default: the answer's single number.
last_value Optional Your last known value for this URL and instruction; big jumps against it are flagged in validation. Default: the service's own history.

Free test requests remaining: 3 of 3.


INPUT PARAMETERS

url
llm_instructions
expiration_seconds
value_field
last_value
API EXAMPLE RESPONSE
JSON
{
    "data": [
        {
            "value": 4066.5
        }
    ],
    "validation": {
        "status": "ok",
        "reason": "no_reference",
        "value": 4066.5,
        "model_value": 4066.5,
        "raw_text": null,
        "suspect_pct": 10,
        "reject_pct": 30,
        "on_page": null,
        "votes": [],
        "mode": "shadow"
    }
}
Extract From Screenshot — CODE SNIPPETS

curl --location --request GET 'https://zylalabs.com/api/14423/ai+web+extraction+api/36428/extract+from+screenshot?url=https://www.investing.com/commodities/gold&llm_instructions=Current gold price in USD as a number, in a field called value&expiration_seconds=600&value_field=value&last_value=4012.5' --header 'Authorization: Bearer YOUR_API_KEY' 


    

API Access Key & Authentication

After signing up, every developer is assigned a personal API access key, a unique combination of letters and digits provided to access to our API endpoint. To authenticate with the AI Web Extraction API simply include your bearer token in the Authorization header.

Headers
Header Description
Authorization Required Should be Bearer access_key. See "Your API Access Key" above when you are subscribed.

No long-term commitment. Upgrade, downgrade, or cancel anytime.

(Save 2 months with annual billing 🎉)

Starts at
$10,000/Year


  • Custom Volume
  • Custom Rate Limit
  • Specialized Customer Support
  • Real-Time API Monitoring

Overview

AI Web Extraction API turns any public web page into structured JSON. Send a URL and describe in plain language what you need ("current gold futures price as a number, in a field called value"); the service fetches the page, prunes it, asks a large language model for exactly those fields and returns them as JSON. What sets it apart is the validation layer. Every numeric answer is checked against the page text (is the number really written on the page, or did the model invent or mis-read it), against decimal-shift patterns (x10, x100) and against the previous accepted value for the same URL and instruction. The result is reported in a validation object with status ok, suspect, rejected or unchecked, the reason, the raw text as written on the page and the thresholds used. The check is informational: you always receive the answer and decide what to do with it. This is the same pipeline that powers Zyla's own market data products on hundreds of thousands of extractions per month. Two endpoints: Extract From Page works on the HTML text of the page and is the fast, cheap default. Extract From Screenshot renders the page, takes a full-page screenshot and asks a vision model, for pages whose data only appears once rendered (charts, widgets, canvas-based tables, heavy JavaScript apps). Both accept an optional cache window (expiration_seconds), an optional last known value to compare against, and the name of the field that holds the number when the answer has several. Typical uses: price monitoring from supplier or exchange pages, competitor tracking, extracting specs or availability from product pages, pulling figures from reports and dashboards, feeding AI agents with verified live numbers, and replacing brittle CSS selectors with one instruction that survives page redesigns.

AI Web Extraction API FAQs

A JSON object with two parts: data, the fields you asked for in your instruction, and validation, a check of the main numeric value (status ok, suspect, rejected or unchecked, the reason, the raw text as written on the page and the thresholds used).

In plain language, through parse_description (text endpoint) or llm_instructions (screenshot endpoint). Name the fields you want back, for example "product name, price in USD as a number and availability, in fields called name, price and in_stock".

Use Extract From Page by default: it is faster and works on the page text. Switch to Extract From Screenshot when the data only shows once the page is rendered, such as charts, widgets, canvas tables or JavaScript-heavy apps.

Whether the number is actually written on the page (on_page), whether it looks like a decimal shift (x10, x100), and how far it moved from the last accepted value for the same URL and instruction. You always receive the answer; the status tells you how much to trust it.

Yes. Send last_value with the last number you stored for that URL and instruction and the validation compares against it instead of the service history. Use value_field when your answer has several numbers.

No. The page and the answer are reused for the same URL and instruction while they are younger than expiration_seconds (default 600 seconds). Send 0 to force a fresh fetch.

Extract From Page reads the server-rendered text. For content that only appears after the page runs its scripts, use Extract From Screenshot, which renders the page in a browser before reading it.

The response comes with error true and a message describing the problem (blocked page, timeout, upstream error). The call is not counted as successful.

Both endpoints return structured JSON data. The Extract From Page endpoint provides fields based on the page's text, while the Extract From Screenshot endpoint returns data derived from rendered visuals, such as charts or widgets.

The key fields include "data," which contains the extracted values, and "validation," which provides the status of the extraction, reason for validation, and checks against historical data.

The response is structured into two main sections: "data," which holds the requested fields, and "validation," which details the accuracy checks performed on numeric values, including status and reason.

Extract From Page is suitable for textual data like prices and specifications, while Extract From Screenshot is ideal for visual data such as charts and dynamic content that requires rendering.

Users can customize requests by specifying the "parse_description" for Extract From Page or "llm_instructions" for Extract From Screenshot, detailing the exact fields they want extracted.

The data is sourced from publicly accessible web pages. The API fetches the content directly from the provided URL, ensuring that the information is current and relevant.

Data accuracy is maintained through a validation layer that checks numeric values against the page text, historical data, and potential decimal shifts, providing users with a trust status for each extraction.

Typical use cases include price monitoring for commodities, competitor analysis, extracting product specifications, and feeding real-time data to AI agents for decision-making.

General FAQs

To obtain your API key, first sign in to your account and navigate to the API you want to use. From the API's Pricing section, choose a plan and complete the subscription process. Once subscribed, return to the API page and you will see your API Access Key displayed at the top of the documentation page. You can use this key to authenticate your requests.

You can’t switch APIs during the free trial. If you subscribe to a different API, your trial will end and the new subscription will start as a paid plan.

The free trial lasts for 7 days and allows you to make up to 50 API requests.

No, the free trial is available only once, so we recommend using it on the API that interests you the most. Most of our APIs offer a free trial, but some may not include this option.

Yes. If the API offers a free trial, you will see a "Free 7-Day Trial" option in its Pricing section. The trial lasts for 7 days and allows up to 50 API requests, enabling you to evaluate the API before subscribing to a paid plan.

Zyla API Hub is like a big store for APIs, where you can find thousands of them all in one place. We also offer dedicated support and real-time monitoring of all APIs. Once you sign up, you can pick and choose which APIs you want to use. Just remember, each API needs its own subscription. But if you subscribe to multiple ones, you'll use the same key for all of them, making things easier for you.

Prices are listed in USD (United States Dollar), EUR (Euro), CAD (Canadian Dollar), AUD (Australian Dollar), and GBP (British Pound). We accept all major debit and credit cards. Our payment system uses the latest security technology and is powered by Stripe, one of the world's most reliable payment companies. If you have any trouble paying by card, just contact us at [email protected]

Additionally, if you already have an active subscription in any of these currencies (USD, EUR, CAD, AUD, GBP), that currency will remain for subsequent subscriptions. You can change the currency at any time as long as you don't have any active subscriptions.
The local currency shown on the pricing page is based on the country of your IP address and is provided for reference only. The actual prices are in USD (United States Dollar). When you make a payment, the charge will appear on your card statement in USD, even if you see the equivalent amount in your local currency on our website. This means you cannot pay directly with your local currency.
Occasionally, a bank may decline the charge due to its fraud protection settings. We suggest reaching out to your bank initially to check if they are blocking our charges. Also, you can access the Billing Portal and change the card associated to make the payment. If these does not work and you need further assistance, please contact our team at [email protected]
Prices are determined by a recurring monthly or yearly subscription, depending on the chosen plan.
API calls are deducted from your plan based on successful requests. Each plan comes with a specific number of calls that you can make per month. Only successful calls, indicated by a Status 200 response, will be counted against your total. This ensures that failed or incomplete requests do not impact your monthly quota.
Zyla API Hub works on a recurring monthly subscription system. Your billing cycle will start the day you purchase one of the paid plans, and it will renew the same day of the next month. So be aware to cancel your subscription beforehand if you want to avoid future charges.
To upgrade your current subscription plan, simply go to the pricing page of the API and select the plan you want to upgrade to. The upgrade will be instant, allowing you to immediately enjoy the features of the new plan. Please note that any remaining calls from your previous plan will not be carried over to the new plan, so be aware of this when upgrading. You will be charged the full amount of the new plan.
To check how many API calls you have left for the current month, refer to the 'X-Zyla-API-Calls-Monthly-Remaining' field in the response header. For example, if your plan allows 1,000 requests per month and you've used 100, this field in the response header will indicate 900 remaining calls.

You can monitor your API usage through the response headers included with every request:

x-zyla-api-calls-monthly-used: Shows the total number of API requests you have used during the current billing period.
x-zyla-api-calls-monthly-remaining: Shows the number of API requests you have remaining for the current billing period.

The 'X-Zyla-RateLimit-Reset' header shows the number of seconds until your rate limit resets. This tells you when your request count will start fresh. For example, if it displays 3,600, it means 3,600 seconds are left until the limit resets.

Yes, you can cancel your subscription at any time. Simply go to the Pricing section of the API you're subscribed to and click the "Unsubscribe" button.

Please note that upgrades, downgrades, and cancellations take effect immediately. Once your subscription is canceled, access to the service will end immediately, regardless of any remaining API calls in your quota.

After 7 days, you will be charged the full amount for the plan you were subscribed to during the trial. Therefore, it's important to cancel before the trial period ends. Refund requests for forgetting to cancel on time are not accepted.
When you subscribe to an API free trial, you can make up to 50 API calls. If you wish to make additional API calls beyond this limit, the API will prompt you to perform an "Start Your Paid Plan." You can find the "Start Your Paid Plan" button in your profile under Subscription -> Choose the API you are subscribed to -> Pricing tab.
You can contact us through our chat channel to receive immediate assistance. We are always online from 8 am to 5 pm (EST). If you reach us after that time, we will get back to you as soon as possible. Additionally, you can contact us via email at [email protected]

Please have a look at our Refund Policy: https://zylalabs.com/terms#refund


Related APIs