Keenable raises $26M from Accel and Conviction. Techcrunch
Keenable raises $26M. Techcrunch
Web Search Infrastructure
for AI Labs and Agents
Web Search
Infrastructure
for AI Labs and
Agents
Web Search
Infrastructure
for AI Labs and
Agents
100B+ documents · <250ms p95 (US East) · from $1 / 1K requests @ 100 RPS+
In production at several AI labs
and inference providers
SOTA continuously learning
at AI scale
SOTA continuously
learning
at AI scale
web search
web search
Overall quality 7-day mean share of ultimate
Overall quality 7-day mean share of ultimate
serper
keenable
parallel-turbo
perplexity
exa
tavily
parallel
$0
$1
$2
$3
$4
$5
$6
$7
$8
80%
70%
60%
50%
40%
Cost, $ per 1K queries Lowest public price
Figure 1. Overall result quality versus cost across search providers.
Quality on the NEEDLE benchmark is reported as the 7-day mean fraction of “ultimate” (oracle) performance across all benchmarks, where ultimate performance is defined by pooling results from all providers and using the judge to select the best possible ranking.
Cost is each provider’s lowest publicly available price per 1,000 queries.
Figure 1. Overall result quality versus cost across search providers.
Quality on the NEEDLE benchmark is reported as the 7-day mean fraction of “ultimate” (oracle) performance across all benchmarks, where ultimate performance is defined by pooling results from all providers and using the judge to select the best possible ranking.
Cost is each provider’s lowest publicly available price per 1,000 queries.
Team that already built
Team that already built
Team that already built
Andrey Styskin
Co-founder & CEO
ex-CEO of Search and Ads
at Yandex: $2B business,
beat Google Search in Russia.
Director of Web Infra
at Amazon AGI.
ex-CEO of Search and Ads at
Yandex: $2B business, beat Google Search in Russia.
Director of Web Infra at Amazon AGI.
ex- CEO of Search and Ads
at Yandex: $2B business, beat Google Search in Russia.
Director of Web Infra
at Amazon AGI.
Matthias Petri
Co-founder & Chief Scientist
ex-Principal Applied Scientist
at Amazon AGI, built the web grounding service behind Alexa.
ex-Principal Applied Scientist at
Amazon AGI, built the web grounding service behind Alexa.
Architect of the trillion-token index behind Amazon-scale retrieval.
ex- Principal Applied Scientist at Amazon AGI, built the web grounding service behind Alexa.
search twice
search twice
One index, three
One index, three
One index,
three
One index, three
products
products
products
1
Search API
for your agents
Search API for your agents
Search API
for your agents
Search
API for your agents
Agent Builder Tier
Pay as you go
For Agent builders and Enthusiasts
Cloud-only access
$4 / 1000 requests
Agent Builder Tier
Pay as you go
For Agent builders and
Enthusiasts
Cloud-only access
$4 / 1000 requests
Agent Builder Tier
Pay as you go
For Agent builders and Enthusiasts
Cloud-only access
$4 / 1000 requests
Frontier Tier
Dedicated capacity for AI scale
For AI Labs and Inference platforms
Cloud & On-premises
$1 / 1000 requests
at 100 RPS+
Frontier Tier
Dedicated capacity for AI scale
For AI Labs and
Inference platforms
Cloud & On-premises
$1 / 1000 requests
at 100 RPS+
Frontier Tier
Dedicated capacity for AI scale
For AI Labs and Inference platforms
Cloud & On-premises
$1 / 1000 requests
at 100 RPS+
2
SELECT
SELECT
Thousands of web pages become a table of values, and the answer is computed over its rows, each one naming the page it came from.
Thousands of web pages become a table of values, and the answer is computed over its rows, each one naming the page it came from.
Prompt
Track researcher moves between frontier labs — who left where, when — with a source on every move.
Web Query Language
SELECT
SEM_NORM(SEM_EXTRACT(content,'researcher who changed labs')) AS researcher,
SEM_EXTRACT(content,
'from lab'),
SEM_EXTRACT(content,'to lab'),
SEM_EXTRACT(content,'date'),
ANY_VALUE(url) AS source
FROM
WEB_SEARCH(8 diverse queries)
WHERE
SEM_MATCH(content,'a researcher moving between labs') GROUP BY researcher;
46 researcher moves between 11 frontier foundation-model labs, Jan 2025 — Aug 2026
Read report
Prompt
Track researcher moves between frontier labs — who left where, when — with a source on every move.
Web Query Language
SELECT
SEM_NORM(SEM_EXTRACT(content,
'researcher who changed labs')) AS researcher,
SEM_EXTRACT(content,
'from lab'),
SEM_EXTRACT(content,'to lab'),
SEM_EXTRACT(content,'date'),
ANY_VALUE(url) AS source
FROM
WEB_SEARCH(8 diverse queries)
WHERE
SEM_MATCH(content,'a researcher moving between labs') GROUP BY researcher;
46 researcher moves between 11 frontier foundation-model labs, Jan 2025 — Aug 2026
Read report
Prompt
Track researcher moves between frontier labs — who left where, when — with a source on every move.
Web Query Language
SELECT
SEM_NORM(SEM_EXTRACT(content,'researcher who changed labs')) AS researcher,
SEM_EXTRACT(content,
'from lab'),
SEM_EXTRACT(content,'to lab'),
SEM_EXTRACT(content,'date'),
ANY_VALUE(url) AS source
FROM
WEB_SEARCH(8 diverse queries)
WHERE
SEM_MATCH(content,'a researcher moving between labs') GROUP BY researcher;
46 researcher moves between 11 frontier foundation-model labs, Jan 2025 — Aug 2026
Read report
Prompt
Track researcher moves between frontier labs — who left where, when — with a source on every move.
Web Query Language
SELECT
SEM_NORM(SEM_EXTRACT(content,'researcher who changed labs')) AS researcher,
SEM_EXTRACT(content,
'from lab'),
SEM_EXTRACT(content,'to lab'),
SEM_EXTRACT(content,'date'),
ANY_VALUE(url) AS source
FROM
WEB_SEARCH(8 diverse queries)
WHERE
SEM_MATCH(content,'a researcher moving between labs') GROUP BY researcher;
46 researcher moves between 11 frontier foundation-model labs, Jan 2025 — Aug 2026
Read report
3
Time Machine
Time Machine
A point-in-time search API that enables you to search across prior versions of web pages and retrieve content from specific historical snapshots. Add query_time and both the corpus and the ranking rewind.
A point-in-time search API that enables you to search across prior versions of web pages and retrieve content from specific historical snapshots. Add query_time and both the corpus and the ranking rewind.
Figure 2. The web changes every minute. Here is one query run at 50 query times across 37 days, from the semifinals to just after the final.
Each dot is a top-10 result at its acquired time, coloured by finalist; the line is query_time advancing. Watch Arsenal appear first as the web tips them for the final, then PSG, then the whole set flip to the result the moment the line crosses kick-off.
Get Independent
Web Search infrastructure
for AI Labs and Agents
Get Independent
Web Search infrastructure
for AI Labs and Agents
Get Independent
Web Search infrastructure for AI Labs and Agents