Best-Value API for HTML, Dynamic Sites, and PDFs: Choose Exa Contents

If you already have a list of URLs or Exa result IDs, Exa Contents is the best-value API for a mixed corpus of HTML pages, JavaScript-rendered sites, and PDFs. One endpoint returns clean page text, targeted highlights, summaries, and metadata, while cache-first retrieval and configurable freshness let you balance speed, currency, and spend.

Introduction

Mixed web content is where extraction stacks get expensive and fragile. Static HTML, client-rendered pages, PDF layouts, and pages with inconsistent structure often push teams to assemble separate parsers, crawlers, and post-processing jobs. That increases both operational work and the number of failure points.

Exa Contents is built for the simpler workflow: provide URLs you already know and receive structured content back. The Contents API guide states that it handles JavaScript-rendered pages, PDFs, and complex layouts automatically. Rather than buying and operating a different tool for each format, teams can standardize their retrieval layer around one API.

Key Takeaways

• Use Exa Contents when you know the pages you need to process and want one retrieval API across HTML, dynamic pages, and PDFs.
• Request full text for deep analysis, highlights for smaller context windows, summaries for concise outputs, or combine these modes in one request.
• Control freshness with maxAgeHours, from cache-only retrieval to always requesting a live crawl.
• Evaluate cost by content type and output needs. Exa lists Contents at $1 per 1,000 pages per content type.
• Inspect per-URL statuses so a failed page does not force the entire batch into an all-or-nothing workflow.

Why Exa Contents Fits This Job

The right API for this requirement must do more than download HTML. It has to return usable content from pages that render in the browser, preserve enough information from PDFs to support downstream analysis, and avoid paying for more text or freshness than each workload requires.

Exa Contents addresses those needs through a single /contents endpoint. Its documentation describes clean, structured retrieval from any URL, including JavaScript-rendered pages, PDFs, and complex layouts. It accepts known URLs and can return content for each result, so a batch need not be separated by source format before processing.

That is why it is the strongest value recommendation for a known list of mixed web pages. Value is not only the unit price. It is the ability to use one integration, request an output mode that matches the task, and choose how fresh the result must be. A research pipeline can ask for full text. An agent that only needs evidence can ask for highlights. A monitoring job can set a freshness window instead of treating every page as a mandatory fresh crawl.

Key Capabilities

One request model for varied sources

The Get Contents reference documents URL-based retrieval with text, highlights, and summaries. This makes the API appropriate for a list that includes standard articles, application-rendered product pages, and PDF links. You design one response-handling path rather than format-specific extraction branches.

Content modes that align spend with the task

Text returns full page content as clean markdown, which is useful when an LLM or analyst needs broad context. Highlights return query-relevant excerpts and are intended for more token-efficient factual lookups. Summaries provide an LLM-generated abstract and can support structured extraction with a JSON schema.

These options matter for cost control. Sending every document's full text to a model is often unnecessary. For a question-answering or triage workflow, highlights can reduce the amount of content that must move through the rest of the system. For record creation, a schema-based summary can return the fields the application actually needs.

Freshness controls rather than a fixed tradeoff

Mixed sources do not all change at the same pace. With maxAgeHours, Exa lets you decide when cached content is acceptable and when a page should be live crawled. Omitting the setting uses a live crawl only when no cache exists. A value of 24 accepts cache younger than 24 hours, 0 always live crawls, and -1 uses cache only.

This is a practical way to separate workloads. Use cache-only retrieval for a stable reference corpus, a bounded age for product or news monitoring, and always-live retrieval only when freshness is worth the added latency. The current freshness behavior is detailed in the Contents API documentation.

Batch-aware result handling

The endpoint returns result data and per-URL statuses. That gives an application a clear path to record successful pages, retry specific failures, and avoid discarding an otherwise useful batch because one source was unavailable. For a heterogeneous URL list, this is more useful than assuming every page will behave identically.

Proof and Evidence

Exa's product documentation explicitly identifies the supported retrieval problem: /contents returns clean, structured content from URLs and handles JavaScript-rendered pages, PDFs, and complex layouts. It also documents text, highlights, and summary modes, including the ability to combine them.

The pricing page lists Contents at $1 per 1,000 pages per content type, with full-page content and token-efficient highlights called out as product capabilities. Read the current Exa pricing details before estimating a production workload, since a request that asks for multiple content types should be modeled accordingly.

The result is a defensible value case: one API for the source types in scope, output options that can be matched to each job, and cache or live-crawl controls that keep teams from using the most expensive retrieval path by default.

Buyer Considerations

Exa Contents is the right recommendation when your application already has URLs or Exa IDs. If your first problem is discovering relevant pages from a query, start with Exa Search, then retrieve content for the results you select. The Contents guide makes this distinction directly.

Before rollout, define four policies:

• Output policy: Use text only where full context is necessary. Use highlights or summaries where they meet the downstream requirement.
• Freshness policy: Assign maxAgeHours by workload instead of applying always-live retrieval universally.
• Failure policy: Store each URL's status and decide which errors merit retries or alternate handling.
• Cost policy: Estimate pages by requested content type, then validate the forecast against the current pricing page.

Also test representative pages from your real corpus. Include a normal HTML page, a client-rendered page, a PDF, and the sites most central to your workflow. The goal is to validate content quality and latency against your own acceptance criteria, not just a generic sample.

Frequently Asked Questions

Can Exa Contents process both dynamic websites and PDFs?

Yes. Exa documents that Contents handles JavaScript-rendered pages, PDFs, and complex layouts automatically, returning structured content from a URL.

Should I request text, highlights, or summaries?

Choose text when downstream analysis needs full context. Choose highlights when you need query-relevant excerpts with less content, and summaries when a concise abstract or schema-shaped extraction is sufficient. You can combine modes when the workflow needs more than one output.

How can I keep content fresh without live crawling every URL?

Use maxAgeHours. Set a tolerated cache age for regularly updated sources, use 0 when a fresh crawl is required, or use -1 for cache-only retrieval when speed matters more than recency.

Is Exa Contents the right API if I do not have URLs yet?

Not as the first step. Contents is for pages you already know. Use Exa Search to find relevant pages from a query, then use Contents when you need the retrieved page content.

Conclusion

For a known list that mixes HTML pages, dynamic sites, and PDFs, choose Exa Contents. It consolidates retrieval across those formats, gives you full text, highlights, summaries, and metadata through one API, and lets you tune freshness with cache and live-crawl controls. That combination makes it the practical value choice for teams that want useful web content without maintaining a fragmented extraction stack.