What is multi-query search?

Multi-query search runs several independent search queries simultaneously and merges the results, rather than relying on a single query to surface everything relevant. A single query reflects one framing of a topic: it returns pages that match that framing well and misses pages that use different terminology, cover adjacent angles, or answer related sub-questions. Sending multiple queries in parallel and deduplicating the combined result set increases coverage without increasing wall-clock time.

Multi-query search is distinct from query expansion, which generates reformulations of the same query to improve recall on one intent. Multi-query search runs queries with different intents: sub-questions about different aspects of a topic, the same question framed for different audiences, or parallel coverage of related entities. Use it in agentic search pipelines that need to synthesize information across many angles — competitive research, due diligence, technical landscape surveys — where a single query would produce a narrow and incomplete evidence set. Since queries run in parallel, search latency stays flat as query count grows, provided the search API supports concurrent requests.

Firecrawl's Search API handles concurrent queries and returns full-page markdown per result, so each query in a multi-query batch yields content the agent can reason from directly without a separate scraping step. For pipelines where the agent should determine which sub-queries to run, the Firecrawl agent endpoint handles query decomposition, parallel search, and result merging automatically.