Keenable Raises $26M Seed to Build Web Search for AI Agents

Keenable was founded by Andrey Styskin and Matthias Petri. Styskin led search, AI, and cloud at the Russian search giant Yandex, and later worked with Petri at Amazon building the web search infrastructure behind AI services like Alexa. Petri is described as a German AI scientist. The two say that while at Amazon, they noticed AI crawlers making up an ever-larger share of search traffic — based on data Cloudflare had published — and concluded that the web needed search infrastructure designed from the ground up with AI in mind, rather than retrofitted for it.

100 billion documents, and a product still on the way

Keenable has built a web search index covering more than 100 billion documents, and the company says its API is already being used in production by several AI labs and inference providers, for both training and serving. It hasn't disclosed specific customer names. More recently, it announced a partnership with Gradium to support real-time information retrieval.

The company's next product is something it calls the Web Query Language — a tool meant to help AI systems pull together answers from multiple sources when no single web page has the full picture. Styskin put it bluntly: "Don't ask, it's painfully expensive," referring to the cost of serving queries across the entire web without tuning the index structure for specific tasks. He says that without such tuning, serving costs balloon to unmanageable levels. The team currently has 15 engineers split across the US and Europe, and the new funding will go toward doubling headcount by year's end and building out a sales and marketing organization.

Where Google and Microsoft are closing the door

Part of what makes this business viable is a shift happening at the big tech companies. Both Google and Microsoft are winding down their legacy search APIs, reportedly to avoid having their own search products cannibalize their AI offerings. Instead, according to Accel partner Zeya Loginov, the two companies are bundling search into other products and becoming much more selective about partners. Styskin sees this as validation rather than a threat — he acknowledges it's still hard to pull people away from Google search itself, but says that when it comes to queries generated by agents, Google is a competitor Keenable can actually beat.

The market already has established players like Brave and Exa. In a "Search Index" benchmark published by Artificial Analysis on August 18, Keenable was evaluated alongside Parallel, Exa, Firecrawl, You.com, Tavily, and Brave — seven search APIs in total. The test ran the same GPT-5.6 Luna model while swapping out only the search provider, with Parallel, Exa, and Firecrawl taking the top spots at scores of 75, 74, and 73 respectively. Amazon entered the same market on August 4, launching its own web-index-based search feature in Bedrock.

Provider	Role in this story
Keenable	$26M seed, index of 100B+ documents
Brave / Exa	Established competitors in agent-focused search
Google / Microsoft	Winding down legacy search APIs, shifting to bundled partnerships
Amazon	Launched its own web-index search feature in Bedrock

Editor's take

There's one conviction running through this story: the belief that the era of "ten blue links" is ending isn't just something startups believe — it's shared inside Google too. The fact that Google itself is winding down its search API and rebuilding the search experience for AI use gives challengers like Keenable an opening. It's a pattern that's played out before in search: when a big company locks the door to protect itself from cannibalization, smaller players find the gap and slip through.

Looking at it generationally makes the shift clearer. AI services used to have no choice but to rent Google's or Bing's API as-is to get search results. Now, companies are building indexes from scratch, designed specifically around how agents query information — and that's giving developers more options for where to source their search grounding. Amazon adding its own index to Bedrock and Keenable building a standalone index sit on the same trend line, but they point in different directions. Amazon is selling within its own ecosystem; Keenable chose to open its API neutrally to multiple AI labs.

For teams building agents, the practical takeaway is that which search API you plug in has become a variable that directly shapes both response quality and cost. As the Artificial Analysis benchmark showed, better search quality means agents burn fewer tokens overall, which lowers total cost. When picking a search provider, it's worth looking past speed and pricing alone to consider how much it actually reduces the need for follow-up queries in real tasks.

Over the next few months, the balance of this market will likely come down to how fast Google and Microsoft close off their APIs versus how quickly alternatives like Keenable, Exa, and Brave can land customers. Keenable has said it plans to double its headcount by year's end, so the next update worth watching for is probably the moment specific customer names start to surface.