Results at a glance
| Feature | Result | What we observed | Run cost |
|---|---|---|---|
| Scrape · exact URLs | 308/343 | Expected website details found | $0.96 |
| Crawl · homepages | 157/343 | Expected website details found | $4.70 |
| Company enrichment | 1/2 | Cloudflare matched. GitHub misidentified. | 4 credits |
| Named people | 2/2 | Correct executives and roles | 0 credits |
| People by role | 1 of 2 matched | CEO first, plus a CEO’s office profile | 4 credits |
| Alexandria | 3/4 | Exact npm version. Release date missing. | 5 credits |
| Search | #1 of 3 | Official React release page | 2 credits |
| Map | #2 of 5 | Official React release page | Unavailable |
| Browser interaction | Passed | One click revealed hidden text | Unavailable |
| Agent | 16/16 | Facts with supporting citations | $0.00 |
| Structured Scrape | 14/16 | Correct facts from exact URLs | $0.075 |
Costs are for the tested runs. Dollar amounts come from AgentRouter receipts. Credit counts come from the Firecrawl plugin.
Website extraction: exact URLs vs. homepages
Both runs come from our November 15 AgentRouter benchmark. We scored the same 343 expected details under two starting conditions.
| Measure | Scrape · exact URLs | Crawl · homepages |
|---|---|---|
| What we supplied | 192 specific page URLs | 100 company homepages |
| How Firecrawl found pages | Read the supplied pages | Chose up to 10 pages per site within the 90-second test limit |
| Expected details found | 308/343 (89.8%) | 157/343 (45.8%) |
| Prices with billing units | 49/59 | 3/59 |
| Median / P95 time | 2.3 s / 5.0 s | 16.6 s / 31.6 s |
| Total run cost | $0.96 | $4.70 |
At Momentic, Scrape found all seven prices from the exact URL. Crawl found zero from the homepage and spent nine of its ten pages on docs. Check which URLs Crawl visited before treating a missing price as absent from the site.
Company enrichment
- Company match
- 1/2Domains resolved correctly
- Source URL
- 0/2Records included a source URL
- Test batch
- 4 creditsThe Companies API
| Input | Company returned | Founded / headquarters | Provider time |
|---|---|---|---|
| cloudflare.com | Cloudflare · correct | 2009 / San Francisco | 351 ms |
| github.com | Dependabot · wrong entity | 2017 / San Francisco | 2,029 ms |
The returned domain matched both inputs, including the wrong entity. Check the company name before writing firmographics into a CRM.
People lookup by name
Given name and employer, DataLegion returned the right person, company and title for both executives. The batch cost 0 credits.
| Input | Returned title | Provider time |
|---|---|---|
| Matthew Prince + Cloudflare | Chief Executive Officer & Co-Founder | 230 ms |
| Satya Nadella + Microsoft | Chairman and Chief Executive Officer | 289 ms |
People search by role
One FullEnrich query for CEO at cloudflare.com returned two profiles. One was Cloudflare’s CEO. The query cost 4 credits and the provider reported 1,178 ms.
| Rank | Person | Returned title | CEO match |
|---|---|---|---|
| 1 | Matthew Prince | CEO & Co-Founder | Yes |
| 2 | Name withheld | Executive Office of the Co-Founder & CEO, Matthew Prince | No · executive office |
The second profile’s title mentions the CEO but describes someone in the executive office. Check the full returned title before saving a contact as CEO.
Alexandria: exact package lookup
- Requested items
- 3/4Name, version and registry URL. Date missing.
- Discovery
- 0 creditsTwo calls to find and inspect the npm capability
- Execution
- 5 creditsOne npm lookup for react@19.0.0
This was one exact-version lookup. If you need the release date, check the publisher’s announcement.
Search and Map
- Search · React 19 release query
- #1 of 3Official page found · 2 credits
- Map · react.dev
- #2 of 5Official page found · cost unavailable
Both found the known announcement. Search used a dated query restricted to react.dev. Map used the domain and “react-19” as a keyword.
Browser interaction
- Before clicking Start
- Hidden“Hello World!” was in the DOM
- After clicking Start
- VisibleFirecrawl saw “Hello World!” on the page
- Session cleanup
- ClosedStop call confirmed session ended
Passed one click-and-wait test on a public demo. Per-call time and charge were unavailable.
Structured answers and citations
- Agent
- 16/16Facts with supporting official citations · $0.00
- Structured Scrape
- 14/16Correct facts · $0.075
- Structured Scrape citations
- 5/16Facts with citations to the supplied page
The pilot covered three questions. Agent started from homepages and could cite any supporting official page. Scrape received exact URLs and its citations had to point to those pages.
Test scope
The four company, people and Alexandria checks reported 13 credits. The 55-second ChatGPT turn includes orchestration. The per-record times above are provider times. Website extraction and the structured-answer pilot use the November 15 AgentRouter results.
Untested in this review: Parse, Monitor, standalone Batch Scrape, developer, research, government and jobs workflows, and contact discovery or verification.
Test data
- Firecrawl plugin runs · November 17 (JSON)
Test inputs and settings, selected responses, providers, timing and reported credits.
- Website extraction runs · November 15 (JSON)
Exact URLs, homepage inputs, crawl limits and per-run results across the tools in the comparison.
- Website extraction scorecard · November 15 (JSON)
Expected field counts, coverage, timing and cost by tool, including both Firecrawl runs.
- Structured answers pilot · November 15 (JSON)
Three questions, API settings, returned facts, citation checks and gateway receipts.