Keenable has emerged from stealth with $26 million in seed funding led by Accel to build independent web search infrastructure designed specifically for artificial intelligence models and agents. Conviction also participated in the financing alongside angel investors from Amazon, ClickHouse, Databricks, Google, Snowflake and SpaceXAI.
The San Francisco-based company is building a search platform intended to give AI systems continuous, cost-efficient access to live information across the web rather than forcing models to rely primarily on information learned during training.
At the center of Keenable’s platform is an independent search index containing more than 100 billion documents.
The infrastructure was designed specifically for how AI agents retrieve and reason over information rather than around the traditional search experience built for human users.
Keenable argues that conventional web search infrastructure creates two significant problems for AI applications: access and economics.
Traditional search products were generally designed around occasional human queries and presenting users with a list of links.
AI agents behave differently.
An agent performing a complex task can repeatedly search for new information, analyze the results, and generate additional queries as its work progresses.
That process can create substantially greater search volume than a traditional user query, making conventional search and page-fetching APIs increasingly expensive when deployed continuously at AI scale.
Keenable also sees greater demand for independent search infrastructure following restrictions on third-party access to search APIs from major technology providers including Google and Microsoft.
The company believes these limitations could make it harder for AI developers to build systems that can continuously retrieve fresh information from across the web.
Keenable’s infrastructure is designed to make those searches an order of magnitude more cost-efficient than traditional search infrastructure, according to the company.
That could enable AI systems to conduct substantially more searches during complex tasks without creating prohibitive infrastructure expenses or latency.
Keenable is also developing new retrieval technologies on top of its underlying search index.
One of its first products is Web Query Language for AI, which is designed to allow models to retrieve, combine and reason over information from thousands of sources across the live web.
The system targets questions and tasks where the necessary information cannot be found within a single webpage or structured database.
Rather than simply retrieving individual pages, Web Query Language is intended to help AI systems assemble information across multiple sources before reasoning over the combined material.
Keenable plans to publicly introduce Web Query Language and several additional products over the coming week.
The company was founded in 2025 by Andrey Styskin and Matthias Petri, both of whom previously developed large-scale search infrastructure.
Styskin previously served as CEO of Yandex Search and later as a director at Amazon AGI.
Petri previously served as a Principal Applied Scientist at Amazon AGI.
Keenable represents the founders’ third effort building a web-scale search index.
At Yandex, Styskin helped build a web index containing approximately 200 billion documents as the company’s search engine became the leading search service in Russia.
Styskin and Petri later worked on large-scale web search infrastructure at Amazon AGI that was used internally across Amazon.
Keenable has already secured commercial contracts with multiple AI labs.
The company is also working with model developers and inference providers to connect AI systems with live web information during both model runtime and training.
The broader thesis behind Keenable is that improving access to external knowledge could become an important avenue for increasing AI performance as models become more capable.
Rather than requiring a model to encode all relevant information into its parameters, Keenable wants AI systems to continuously retrieve higher-quality and more current knowledge as they work.
The new funding will support expansion of Keenable’s engineering organization across the Bay Area and Europe.
Capital will also be used to expand the company’s independent web index and crawling infrastructure and develop the retrieval technologies underlying Web Query Language.
KEY QUOTES:
“Today’s leading AI models are excellent summarisation machines. They can answer almost any question, but ask them about a subject you know deeply, and you’ll quickly find their limits. The next frontier, that will take average answers to outstanding ones, will come from giving AI dramatically better access to the world’s knowledge. Our goal is to make it accessible at AI scale.”
Andrey Styskin, Co-Founder And CEO Of Keenable
“Search is becoming foundational infrastructure for the next generation of AI, but today’s systems were never designed for the way models access and use information. Andrey and Matthias have already built web-scale search systems at two of the world’s leading technology companies, making them uniquely positioned to rethink search for the AI era. We’re excited to support them as Keenable builds the knowledge foundation for the next frontier.”
Zhenya Loginov, Partner At Accel
