Which API fits the job?
Start with what you already know. If you have the target URLs, test content extraction directly. If you need to discover sources, compare Search. If you need a whole site, define crawl coverage and failure criteria instead of reusing a search benchmark.
When to consider Exa
Consider Exa when discovery and selective reading are the core loop. Its separate Contents endpoint lets you make the reading step explicit rather than fetching every result by default.
When to consider Firecrawl
Consider Firecrawl when the job includes scraping, mapping or crawling sites after discovery. Price the actual pages and search credits consumed rather than treating every operation as one equivalent request.
What is measured and what is documented?
This page compares checked documentation and product contracts. We have not completed a matched All Them APIs run for this pair and assign no synthetic winner or composite score.
External evaluations can help choose a first configuration, but they use different models, topics, result limits and cost definitions. Review the source register before importing a ranking into your own decision.
Request shape, pricing and agent integration
| Decision | Exa | Firecrawl |
|---|---|---|
| Search job | Ranked retrieval with selectable modes and content fields | Search for URLs, optionally followed by scraping |
| Known URLs | Contents for supplied URLs | Scrape for one page; Map/Crawl for site workflows |
| Billing shape | Mode-based request pricing with possible content add-ons | Plan credits; search and page-reading workloads consume different units |
| Free evaluation | Monthly Exa credit allowance | Published Firecrawl free plan credits with concurrency limits |
| Our source-retrieval report | Available, August 23 | No equivalent All Them APIs run yet |
Prices are published USD list rates or explicitly labeled credit calculations, not observed costs from a new test. Extra results, content options, model tokens, account limits and minimum commitments can change the total. Check the linked source for the exact configuration.
A useful evaluation before switching
- Score source discovery separately from successful page extraction. A returned URL can still fail when the agent tries to read it.
- For a crawl, name the required pages and field completeness before the run. A fast homepage scrape does not establish site coverage.
- Keep JavaScript rendering, freshness/cache policy and content limits in the record. Different retrieval settings can dominate both price and latency.
Keep the same acceptance criteria across providers and save failures as failures. Avoid changing the target source, query or score threshold after seeing which vendor wins. Measure both useful evidence and the total workflow needed to obtain it.
Sources and review dates
Documentation review: October 27, 2026. Our own measured retrieval runs: August 23, 2026. External reports retain their publisher and measurement window.
Corrections: email faizank@docsalot.dev with an official source.