Firecrawl search endpoint: limit 100 vs Sume's limit of 10

Sume's search endpoint is POST /v1/firecrawl/search. It takes a required query (up to 500 characters) and an optional limit from 1 to 10, with 5 as the default. It returns URLs, titles and snippets only, so a search is a way to pick pages, not evidence you can cite yet.

Firecrawl numbers are from its search reference and Sume's from its OpenAPI file and MCP tool descriptions, all read 2026-09-30.

How do the two search endpoints differ?

Firecrawl's reference lists a default limit of 10 and a maximum of 100, sources of web, images, news and alexandria, a timeout and a tbs date filter such as qdr:w. Sume's request schema rejects unknown properties, so none of those are accepted.

What does a Sume search return?

The MCP description for crawl_search says it is for discovering public sources from a query when you have no known URL. It returns URLs, titles and snippets, not full evidence. Use crawl_map to search a known site's inventory, and crawl_scrape to read a known page.

const res = await fetch("https://api.sume.com/v1/firecrawl/search", {
  method: "POST",
  headers: {
    Authorization: "Bearer " + process.env.SUME_API_KEY,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({ query: "mcp streamable http transport", limit: 10 }),
});
console.log(res.status, await res.json());

What if I need more than 10 results?

Rephrase the query. Several narrower queries with distinct terms give more distinct pages than one wide query, and each call stays inside the 10-result ceiling. If you already know the site, call crawl_map (limit 1-100) to list its pages.

What is the next step after a search?

The tool description says to crawl_scrape the selected sources and cite them. Pick a few URLs from the snippets, scrape each, and quote from the scraped markdown rather than from the snippet. Options for that call are in Firecrawl scrape formats on Sume. The Crawl group is listed in MCP tools and gates.