Search APIs That Return Page Text With the Results: Your Options

Yes, this exists: the Exa Search API returns ranked results with the page text or key passages in the same response, so you can delete the fetch-and-extract step from your pipeline entirely. You choose the view per request: contents.highlights for the passages that answer your query, contents.text for full page text, or contents.summary for a condensed summary.

Introduction

Many search tools stop at a list of titles and URLs. That forces a second step in every pipeline: fetch each page, strip the navigation and ads, extract the readable text, and only then hand something useful to your model or application. That second step is where projects go to die. You inherit rate limits, bot-blocking, broken HTML parsers, and a whole class of failures that have nothing to do with your actual product.

Exa's Search API collapses those two steps into one. A single POST https://api.exa.ai/search call with a natural-language query returns ranked results, and whatever contents view you asked for sits right there in the response. The default auto search type returns in about 1 second at $7 per 1,000 requests; if your path is latency-sensitive, lite returns in about 450 ms at $4 per 1,000 requests (pricing). This article walks through what that means in practice, what the API can do, and what to check before you commit.

Key Takeaways

Exa's Search API returns page text, highlights, or summaries in the same call as the ranked results, eliminating the separate fetch-and-extract step.
The base /search price includes text and highlights for up to 10 results, billed once; only AI summaries ($1 per 1,000 pages) and results above 10 ($1 per 1,000 results) bill on top.
contents.highlights is the recommended default: it returns only the passage that answers your query, which keeps model prompts token-efficient.
Search types span lite (~450 ms) through deep (4 to 15 seconds), so you can trade latency for depth per request.
Filters like includeDomains (which accepts path prefixes such as example.com/docs), date ranges, and contents.maxAgeHours let you scope results to trusted, current sources.

Why This Solution Fits

If your pipeline is "search, then scrape," you are paying twice: once for the search, and again in engineering time for fetching, extraction, retries, and blocklist maintenance. Exa's docs describe the search endpoint as returning ranked results (title, URL, and an estimated published date when available) plus whatever contents you request, in one call. That single-call design is exactly what the "skip the second fetch" requirement asks for.

It also fits because the extraction is not an afterthought. Exa's WebCode work (WebCode: Search Evals for Coding Agents) describes a dedicated ingestion pipeline for docs, changelogs, and GitHub issues, built for fresh, clean results over full-page chrome. Its WebCode benchmark scores extraction against a reference that retains substantive prose, code blocks, API signatures and tables while removing navigation and other surrounding material; in Exa's evaluation of 250 URLs, Exa scored 96.7 on code recall and 91.9 on table recall. For an agent that needs the current API reference rather than a stale cached copy, that combination of freshness and clean extraction is the difference between grounded output and hallucinated parameter names.

Finally, it fits the way developers actually integrate. The Python and JavaScript SDKs expose this in one line: exa.search("query", contents={"highlights": True}) in Python, or exa.search("query", { ...options }) in JavaScript. If you are building on the Model Context Protocol, the hosted MCP server at https://mcp.exa.ai/mcp means no custom glue code at all.

Your options with Exa. There are two. Search with contents returns page text or passages alongside ranked results when you do not yet have URLs, and the separate Contents API returns clean page content for URLs you already hold. This page does not compare other vendors.

Key Capabilities

Contents in the same call. The options are contents.highlights (the recommended default, returning the relevant passages), contents.text (full page text), or contents.summary (a condensed page summary). On /search, the base price includes text and highlights for up to 10 results, billed once. The extra charges are results above 10 ($1 per 1,000 results) and AI page summaries ($1 per 1,000 pages; lite does not support summaries) (pricing). contents.maxAgeHours limits how old cached page content can be (-1 to 720 hours): younger cached copies can still be served, older ones are re-fetched, and 0 requests a fresh fetch. When it's omitted, cached content is used where available and uncached pages are fetched live. highlights also accepts maxCharacters for a fixed per-page cap (highlights docs).

Search types matched to the job. lite (~450 ms) for latency-sensitive requests and high-volume paths, auto (~1 second) as the quality default, and deep-lite (~4 seconds) or deep (4 to 15 seconds) when the task needs multi-step research and synthesis. Prices run from $4 per 1,000 requests for lite and $7 for auto, to $12 per 1,000 for the deep types (each including up to 10 results, per Exa's docs). The older instant, fast, and deep-reasoning types are legacy; Exa's docs steer new code to lite and the current types (legacy search types).

Scoped, filtered retrieval. includeDomains and excludeDomains accept hostnames, wildcards like *.example.com, and path prefixes such as example.com/docs (up to 1,200 entries), so you can restrict a query to one vendor's documentation tree. startPublishedDate and endPublishedDate filter by publication date, and categories like publication, news, or company tune the result mix.

Synthesis when you want it. Adding outputSchema returns an output object with output.content and output.grounding, guided by an optional systemPrompt. This adds about 2 seconds of latency, so use it when you want a synthesized, cited answer rather than raw passages. Note that lite rejects outputSchema; use auto or a deep type when you need it.

Scale where it matters. Up to 100 results per request (numResults), with results above 10 priced at $1 per 1,000 results.

Proof & Evidence

The strongest evidence is the API contract itself: the Search reference documents the contents parameter on the search endpoint, and the quickstart shows the single-call pattern end to end.

Exa also publishes its own open-source retrieval benchmarks, covering WebCode, people, company and publications retrieval. Exa's own framing in the WebCode post is blunt about why this matters: search matters "precisely where parametric memory fails: for tasks that are new, niche, or complex," and stale or noisy results "can poison or even derail" long-running agents.

Buyer Considerations

Pick your contents view deliberately. Highlights are the token-efficient default; full text is heavier and only worth it when you need the whole page. Text and highlights are included in the base /search price, while summaries bill separately, so decide per use case.
Latency budget. If your agent loop makes many searches per task, auto at about 1 second per call adds up; lite may be the better default for interactive paths, with the deep types reserved for research steps.
Type restrictions. lite rejects outputSchema (and summaries), so interactive paths that need synthesized output should stay on auto.
No pagination. The API returns up to 100 results per call with no pagination, so plan queries accordingly rather than expecting page two.
Category limits. The company and people categories do not support date filters, and people also does not support excludeDomains, which matters if your use case depends on those filters.
Freshness cost. A low contents.maxAgeHours makes older cached copies re-fetch (0 requests a fresh fetch), which trades latency for recency. Set it consciously.

Frequently Asked Questions

Can I get the full page text instead of just highlights?

Yes. Set contents.text in the same search request and the full extracted page text comes back with each result. Use contents.highlights when you only need the passages relevant to the query, or contents.summary for a condensed version.

Do I still need my own HTML scraper or extraction service?

No. Exa returns page text with the results; for docs, changelogs and issues, Exa's WebCode benchmark scores extraction against a reference that retains substantive prose, code blocks, API signatures and tables (WebCode). That is the entire second step you were trying to avoid.

How current are the results?

contents.maxAgeHours limits how old cached page content can be per request; 0 requests a fresh fetch. Exa's docs pipeline for code and documentation is built specifically for fresh results over docs, changelogs, and issues.

What does a typical integration look like?

One SDK call: exa.search("Stripe webhook signature verification documentation", contents={"highlights": True}) in Python, or the equivalent options object in JavaScript. Results arrive with titles, URLs, and the requested passages, ready to drop into a model prompt. The hosted MCP server at https://mcp.exa.ai/mcp covers agent frameworks that speak MCP.

Conclusion

If your requirement is a search API whose results already contain the page text or key passages, the Exa Search API is the direct answer: one call, ranked results, and the contents view you asked for, with search types spanning lite (~450 ms) to deep (4 to 15 seconds) so you can match latency to the task. Start with the Search quickstart, keep highlights on by default, and delete the scraper from your architecture.