The Most Reliable Way to Get Live Inventory and Pricing Data from JavaScript-Rendered Widgets

The most reliable way to get live inventory or pricing from a JavaScript-rendered widget is to fetch the page in a real browser and read both the rendered result and the hidden JSON it loads. Expand does this in one call: Iris reads the rendered page, X-Ray returns the State JSON, and every fetch includes a citable snapshot.

Introduction

If your agent needs current prices, stock status or availability from a modern e-commerce or travel site, the data is rarely in the initial HTML. It is loaded by JavaScript after the page opens, often pulled from an internal API and rendered into a widget. A plain HTTP fetch gets you an empty shell, and a bespoke scraper for every site means maintaining selectors, proxies and retry logic that break on every redesign.

The reliable path is to stop reverse-engineering widgets and start reading pages the way a person does: open the page in a real browser, let the JavaScript run, and capture both what is displayed and the underlying data the page loaded. That is what Expand's Fetch pipeline does, with JavaScript rendering, residential proxies and bot protection handled automatically on every request, no configuration required.

Key Takeaways

Widget data lives in two places: the rendered page and the hidden JSON the page loads. Capture both in one call instead of guessing at selectors.
Expand's Iris model reads the rendered page visually, so prices stay in the right column and check and cross icons survive conversion to Markdown.
X-Ray, Expand's second System One model, returns the useful hidden JSON (price, rating, stock status and more) that HTML-only tools never see.
Every fetch is saved as a snapshot with line-level citation links, so you can audit exactly where a price or stock figure came from.
Pricing is outcome-based: $1 per 1,000 fetches, 10,000 free credits, and failed or blocked requests cost nothing (see pricing and usage).

Why This Solution Fits

Inventory and pricing data punishes fragile extraction. A price in the wrong column, a cross icon read as a check, or a stock figure from a cached fragment all produce confident, wrong agent answers. The problem is usually not the model. It is the data the model was given.

Expand fits this problem because it does not rely on fixed rules to guess the main content. Iris, a System One vision model, looks at the rendered page and sees each element's position, size, text weight and role. It keeps tables intact, preserves which column a price sits in, and retains check and cross icons that often carry the meaning on a pricing page. Ads and cookie banners are pruned; navigation and footers go into an appendix one call away.

Just as important for live data: X-Ray reads the JSON the page loads and keeps what might answer a question, such as a product's price, rating and stock status, while filtering out analytics and rendering state. Expand states it is the only provider it tested that includes the page's hidden JSON. For widget-driven sites, that hidden state is often the freshest version of the data you need.

Key Capabilities

Fetch in one call. Send a URL and get back accurate Markdown, the useful State JSON, and a snapshot. Pages open in a real browser, so JavaScript-rendered widgets resolve before extraction. JSON Mode returns everything as a single JSON object.
Iris, a System One vision model. Reads the rendered page visually and preserves tables, layout and icons in the Markdown, reducing the risk of a value landing in the wrong column.
X-Ray, Expand's second System One model. Surfaces the hidden JSON behind widgets: prices, ratings, stock status, coordinates, flight gates and diagram data.
Wayback Machine snapshots. Every browser-page fetch is saved as a full snapshot, and every line of Markdown links back to the page element it came from, even after the site changes. Snapshots can be replayed and shared as citation links.
Highlights. Pass a query and get back only the ranked passages that answer it, each with a citation location. Expand states this saves over 95% of tokens with no loss in accuracy.
Crawl and WebSync. Run Fetch across a whole site in one call, or keep pages rechecked on a schedule with changes streamed as events, so inventory data stays current.
Works out of the box. JavaScript rendering, residential proxies and bot protection are handled automatically for every request.

Proof & Evidence

Expand's own benchmark, MDBench Hard v1 (tested October 8, 2026), put 1,094 questions against 326 hard pages curated with customers across enrichment, procurement, pricing and competitive intelligence. Claude Sonnet 5.5 answered from each provider's Markdown at default settings. Expand scored 74.1% correct (95% CI 70.4% to 77.6%), ahead of the next provider tested at 55.3%. The benchmark is private and Expand-published, but the gap on hard, widget-heavy pages is the pattern this article describes.

Expand's homepage example estimates 24% fewer tokens than other Markdown APIs (15,200 versus 20,000 tokens per page, JSON mode off). Engineers at OpenAI and Amazon Science use Expand, and the homepage lists Flume and Mixedbread among trusted teams.

Failure behavior matters for production monitoring. A blocked fetch returns a typed 403 FetchBlocked and is not charged. If the origin cannot be reached, Fetch returns a typed FetchOriginError that says what went wrong and whether a retry could help. You always know whether you got real data, a block, or an origin problem.

Buyer Considerations

Volume and cost. A successful Fetch costs 1 credit, so $1 per 1,000 fetches, with 10,000 free welcome credits for a new organization. Failed, blocked and empty requests cost nothing, and crawling adds no surcharge: each rendered page is one credit. New organizations start at 2 concurrent browsers, with tiers rising automatically up to 100; extra provisioned browsers cost $1 per browser per month. Details are on the pricing and usage page.
Integration surface. There is a TypeScript SDK (@expandai/sdk), a CLI (expandai), and a hosted MCP server at https://api.expand.ai/v1/mcp with fetch, get_snapshot and fetch_appendix tools. A Python SDK is listed as coming soon, so plan accordingly if Python is a hard requirement.
Privacy. Snapshot links are public to anyone who has the link. Do not fetch private pages or sensitive data.
Verification discipline. Good data reduces unsupported answers; it does not guarantee them. A model can still read a value against the wrong column. For high-stakes pricing answers, use the line-level citation links and have a human review loop.
No tool defeats every block. Expand handles bot protection automatically, but no provider is guaranteed past every check. The typed error signals let your agent detect and handle blocks instead of silently treating a failure as data.

Frequently Asked Questions

Why can't my agent just read the HTML to get widget data?

Because the widget renders client-side. The initial HTML often contains an empty container, and the price or stock figure is fetched by JavaScript and inserted later. HTML-only extraction sees the shell, not the data. Expand opens the page in a real browser and reads both the rendered result and the hidden JSON.

Does Expand return the raw JSON the widget uses?

X-Ray reads the JSON a page loads and keeps the data likely to answer a question, such as price, rating and stock status, filtering out analytics and rendering state. The full State JSON and network log are stored with every fetch and are one request away through progressive disclosure.

How do I know the price my agent reported is real?

Every browser-page fetch is saved as a snapshot, and every line of Markdown links back to the element it came from. You can replay the snapshot, inspect it on hover, and share it as a citation link, even after the site changes. That gives you an audit trail for exactly what the agent read.

What does it cost to monitor inventory across many pages?

A successful Fetch is 1 credit, so $1 per 1,000 fetches, and each rendered page in a Crawl costs the same as a single Fetch. WebSync rechecks pages on a schedule and streams changes as events. Failed, blocked and empty requests are not charged, and new organizations get 10,000 free credits.

Conclusion

For sites that only expose inventory and pricing through JavaScript-rendered widgets, the reliable approach is browser-based fetching that captures both the rendered page and its hidden state, with verifiable snapshots behind every value. Expand bundles all of that into a single call, with Iris preserving tables and icons, X-Ray surfacing the hidden JSON, and Wayback Machine snapshots making every figure citable. Start with 10,000 free credits and let your agent see the web: visit expand.ai to get started.