What is the difference between browserfoundry, Keirolabs and Keiro?
They are one company and one API. Keirolabs is the company. browserfoundry is the name it uses for its developer product: the search, extraction, research and browser endpoints that share one key and one credit balance. Keiro is the name attached to the search engine and its deep search endpoint, the one behind the SimpleQA claim. If you sign up on either site you get the same platform, the same plans and the same credits.
How much does browserfoundry cost?
The free tier gives 750 credits every 30 days with no card. Paid plans are $29.99, $49.99 and $99.99 a month, with 11,250, 33,750 and 112,500 credits. Each endpoint draws a different number of credits: lite search is the cheapest at $0.10 per 1,000 on Essential, search with content costs more, and deep search costs 5 credits. Keirolabs lists deep search at $4.44 per 1,000 requests. One-time credit packs last six months.
Is the 95.3% SimpleQA score comparable to Tavily or Exa?
Only roughly. Keirolabs scores retrieval: a question counts if the gold answer text appears anywhere in the returned snippets or pages. The raw substring score was 87.9%, and a hand audit of misses raised it to 95.3%. The Tavily, Exa and Parallel numbers it compares against were produced by other harnesses, where an LLM reads the results and a judge grades the answer. Keirolabs calls the gaps directional, and that is the right way to read them.
Does browserfoundry render JavaScript?
Yes. Pages are fetched and rendered on Keirolabs' side, and deep search reads each candidate page in its final rendered state. For anything a search result can't cover, the browser API leases a real Chrome session with its own fingerprint and a fresh IP, billed only while the session is open. That makes it one of the few search APIs that also sells browser sessions under the same account.
How fast is Keiro deep search?
It is slow next to plain search. In Keirolabs' own SimpleQA report, median latency for deep search was 14.7 seconds, and it read an average of 7.9 pages per question. That is the price of reading every candidate page end to end. For interactive chat, the fast search endpoint, which returns ranked results in about a second, is the better fit. Deep search suits verification steps and research agents that can wait.
What are the best alternatives to browserfoundry?
Tavily is the most common choice for agent search and has the widest framework support. Exa sells its own deep search tier on an embeddings-based index. Parallel offers search, extraction and multi-step research with much larger backing. Pick browserfoundry if price per call and having search plus browsers on one bill matter most, and you are comfortable buying from a small, newer vendor.
