Firecrawl adds Highlights to /scrape with 100x fewer tokens

• In firecrawl's launch post, Firecrawl added a highlights output mode for /scrape that returns matching text, code, or tables for a query instead of a full-page payload.
• According to firecrawl's benchmark post, the company tested Highlights on 10,000 URLs against Exa Highlights using a "Focus" metric that scores how closely a snippet matches an ideal passage.
• The Highlights docs say the format returns selected text in a highlights field, requires a query, and costs 5 credits per page, 1 base credit plus 4 for the LLM call.
• firecrawl's earlier Question launch framed the companion question feature as a way to skip the scrape, chunk, embed, retrieve loop, while the scrape docs expose the same capability under the query format name.
• In firecrawl's PHP SDK post, the company also pushed a PHP and Laravel wrapper, which means the new /scrape modes are already available outside the Python and Node SDK path.

You can check the new Highlights docs, compare them with the older Question docs, and the PHP SDK docs already show composer require firecrawl/firecrawl-php. The interesting bit is how fast Firecrawl is turning /scrape into a menu of targeted outputs: full markdown, structured JSON, grounded answers, and now passage selection sized for agent context windows.

Firecrawl is positioning Highlights as a narrower retrieval primitive than plain scraping. In firecrawl's launch post, the company says agents can pass a URL and a query, then get back the exact text, code, or tables they need with up to 100x fewer tokens.

The docs add the concrete mechanics that the tweet leaves out:

• Highlights docs say highlights returns source text in a highlights response field.
• The same docs require a query inside the format object, with a maximum length of 10,000 characters.
• Highlights is priced at 5 credits per page, 1 base scrape credit plus 4 additional credits for the LLM call.

The benchmark claim is notable mostly because Firecrawl names the comparison set directly. In firecrawl's benchmark post, the company says it scored Highlights against Exa Highlights on a 10,000 URL test set using a "Focus" metric, defined as closeness to an ideal passage on the page.

Two days before Highlights, Firecrawl shipped what the tweets call the Question format. In firecrawl's loop-skipping explainer, the company said agents can pass a URL and a question, then skip the usual scrape, chunk, embed, retrieve loop for tasks like pulling a rate limit from docs or plan details from a pricing page.

The docs use slightly different naming. In the scrape docs, the capability appears as query, with a prompt and optional mode, and the answer comes back in an answer field. That makes the new Highlights mode easier to place: Firecrawl now has one /scrape variant for grounded answers and another for excerpt extraction.

The quieter release in the same window was language coverage. In firecrawl's PHP SDK post, Firecrawl said developers can scrape pages to markdown, run live web searches, and navigate dynamic sites from PHP or Laravel.

The PHP SDK docs confirm the install command from the tweet, composer require firecrawl/firecrawl-php, while the Node SDK docs keep the existing @mendable/firecrawl-js package path. That matters here because the new output formats are landing as /scrape options, not as separate products, so each SDK pickup widens who can use Highlights and Query on day one.