Keenable Exits Stealth With $26M for AI Search Index

Fundraise Insider tracks newly funded startups each week and delivers verified sales leads of C-suite contacts straight to your inbox, so you reach the right people while the funding is still fresh. See pricing.

Keenable has come out of stealth with a $26 million seed round led by Accel. Conviction also took part, alongside angel investors drawn from Amazon, ClickHouse, Databricks, Google, Snowflake and SpaceXAI.

The company works from San Francisco. Its engineering will soon straddle two continents rather than one corner of California. The product is independent web search infrastructure built for machines rather than people. Its aim is to give models continuous and affordable access to live information instead of leaving them to lean on whatever they absorbed during training.

Sitting underneath the platform is an independent index of more than 100 billion documents, designed around how agents retrieve and reason over information rather than around the familiar page of blue links.

Keenable frames the problem as two related failures of the existing stack: access and economics. Traditional search products assume an occasional human query answered with a list of links. An AI agent working through a complex task behaves nothing like that. It searches, reads what comes back, and fires off further queries as its understanding shifts, generating far more volume than any person would. Conventional search and page fetching APIs become expensive fast when run continuously at that scale.

There is a second pressure the company points to. Major providers including Google and Microsoft have tightened third party access to their search APIs, which makes independent infrastructure more valuable to anyone building systems that need fresh information from across the web. Keenable claims its own approach is an order of magnitude more cost efficient, which would let AI systems run far more searches during a task without running into either cost or latency walls.

On top of the index, the company is building retrieval technology of its own. The first product is Web Query Language for AI, meant to let models pull, combine and reason across thousands of sources at once. It is aimed at the class of question where no single webpage or structured database holds the answer, so the system has to assemble the material before it can reason about it. Keenable plans to introduce Web Query Language and several other products publicly over the coming week.

The founders have done this before. Keenable was started in 2025 by Andrey Styskin and Matthias Petri, and it is their third attempt at a web scale search index. Styskin was CEO of Yandex Search, where he helped build an index of roughly 200 billion documents while the engine became the leading search service in Russia, and later served as a director at Amazon AGI. Petri was a Principal Applied Scientist at Amazon AGI, where the two worked together on large scale search infrastructure used internally across the company.

Commercial contracts with several AI labs are already signed, and Keenable is working with model developers and inference providers to wire live web information into systems during both runtime and training. The new money funds engineering hiring across the Bay Area and Europe, a larger index and crawling operation, and further work on the retrieval technology behind Web Query Language.

Companies selling into fast growing AI infrastructure firms can reach decision makers early through B2B software sales leads compiled from fresh funding rounds.