Andrey Styskin raises $26M for Keenable's agent-first web index

Keenable, founded by former Yandex search executive Andrey Styskin and Amazon retrieval scientist Matthias Petri, emerged from stealth with a $26 million seed round led by Accel to build an independent web index for AI agents. The index contains more than 100 billion documents and is already used in production by several AI labs and inference providers, with a Search API offering p95 latency below 250 milliseconds and a $4-per-1,000-request tier for individual builders. The 15-engineer team plans to double headcount by the end of 2026.

Andrey Styskin raises $26M for Keenable's agent-first web index Former Yandex search executive Andrey Styskin and Amazon retrieval scientist Matthias Petri are building Keenable with a 15-engineer team across the US and Europe. By RuntimeWire Staff /author/runtimewire-staff · Published Primary source: TechCrunch https://techcrunch.com/2026/08/25/accel-backed-keenable-is-indexing-the-web-for-ai-agents/ Why it matters Search is becoming a costly infrastructure layer for agents as Google and Microsoft restrict standalone access. Keenable's founders are betting an independent index can win on speed, price and neutrality. Andrey Styskin @styskin https://x.com/styskin?ref=runtimewire and Matthias Petri https://www.amazon.science/author/matthias-petri?ref=runtimewire have brought Keenable https://keenable.ai/?ref=runtimewire out of stealth with a $26 million seed round to build an independent web index for AI agents. Accel led the financing, with participation from Conviction Partners https://www.conviction.com/moe?ref=runtimewire and unnamed angel investors, according to TechCrunch's August 25 report https://techcrunch.com/2026/08/25/accel-backed-keenable-is-indexing-the-web-for-ai-agents/?ref=runtimewire . The founders spent years inside the search operations Keenable now wants to unbundle. Styskin led search, AI and cloud operations at Yandex before becoming director of Web Information Services at Amazon's artificial general intelligence organization. At a 2024 USC-Amazon symposium, he was already presenting on cost-efficient crawling and retrieval as large language models reshaped search. Petri was a principal applied scientist in Amazon's Alexa AI organization, where his research covered inverted indexes, compression and low-latency retrieval. Styskin and Petri later worked together on Amazon web-search infrastructure used by AI applications including Alexa, TechCrunch reported. That experience gave the founders a close view of a basic mismatch: conventional search products are designed to return a short results page to a person, while agents may issue repeated queries, read many documents and combine evidence before producing one answer. Keenable is betting that the mismatch is large enough to support another independent search index, one of the more expensive pieces of infrastructure a seed-stage founder can choose to build. The index is the product Keenable says its index https://keenable.ai/?ref=runtimewire contains more than 100 billion documents and is already used in production by several AI labs and inference providers during model training and runtime. Keenable has not named those customers, and the index-size claim has not been independently verified. The commercial product is a Search API https://keenable.ai/?ref=runtimewire that exposes web-page search and content fetching through REST, command-line and Model Context Protocol interfaces. Keenable advertises p95 latency below 250 milliseconds in the US East region, along with a $4-per-1,000-request tier for individual agent builders. Higher-volume customers are quoted $1 per 1,000 requests at 100 requests per second or more, with cloud and on-premises deployment options. Those numbers explain part of the $26 million seed. Crawling, storing, refreshing and ranking a meaningful portion of the public web is a continuous infrastructure expense. "Don't ask - it is painfully expensive," Styskin told TechCrunch. Keenable currently has 15 engineers distributed across the United States and Europe https://techcrunch.com/2026/08/25/accel-backed-keenable-is-indexing-the-web-for-ai-agents/?ref=runtimewire . Styskin plans to double headcount by the end of 2026 and build a go-to-market operation, according to TechCrunch. The valuation and precise closing date of the financing were not disclosed. Accel partner Zhenya Loginov https://www.accel.com/team/zhenya-loginov?ref=runtimewire led the deal, TechCrunch reported. Keenable is also developing products that depend on owning the underlying corpus. Time Machine, currently offered through early access, lets developers search historical versions of pages and retrieve content from historical snapshots. An upcoming system called WebQueryLanguage is intended to combine information across sources when no single page contains a complete answer. In July, Keenable announced an integration with Gradium's voice-agent framework https://keenable.ai/blog/natural-conversation-live-retrieval-keenable-search-with-gradium?ref=runtimewire , placing live web retrieval inside a spoken conversation. A text agent can conceal a slow lookup behind a progress indicator. A voice agent leaves the user listening to silence. Big platforms created the opening The timing follows a retreat from general-purpose search APIs by the companies with the largest indexes. Microsoft retired its Bing Search APIs https://learn.microsoft.com/en-us/lifecycle/announcements/bing-search-api-retirement?ref=runtimewire in August 2025. Google's Custom Search JSON API https://developers.google.com/custom-search/v1/overview?ref=runtimewire is among the search products whose access has been restricted or changed, while the tech giants are moving toward more selective, bundled offerings. Amazon is moving in the same bundled direction. AWS now offers Web Search on Amazon Bedrock https://aws.amazon.com/blogs/machine-learning/introducing-web-search-on-amazon-bedrock-for-foundation-model-grounding/?ref=runtimewire , giving developers access to web search from inside its model platform without onboarding an outside search vendor. That creates Keenable's opening and its central sales problem. AI developers losing access to broad, standalone APIs need alternatives. Large customers may still prefer search packaged by the cloud provider already handling their models, security reviews and billing. Keenable is selling independence, lower stated costs and retrieval designed for machine workloads against the convenience of an integrated cloud stack. Styskin argues that agents create a different feedback loop from consumer search. Google learned from clicks and other human behavior; agents generate machine-scale query patterns and can evaluate much more of each retrieved document. Keenable's thesis is that a search system tuned around those patterns can narrow the web faster and serve results more cheaply than infrastructure inherited from consumer search. Search infrastructure is drawing serious capital Keenable enters a category where investors have already committed far larger sums. Exa raised $250 million at a $2.2 billion valuation https://exa.ai/blog/announcing-series-c?ref=runtimewire in May, saying its crawlers track more than 500 billion URLs. In February, Nebius agreed to acquire Tavily https://nebius.com/newsroom/nebius-announces-agreement-to-acquire-tavily-to-add-agentic-search-to-its-ai-cloud-platform?ref=runtimewire , adding real-time search to an AI cloud platform. Brave https://brave.com/search/api/?ref=runtimewire also sells access to its independent index. The spending reflects the economics of agent software. A model can be swapped through an API call, while reliable access to current web information requires crawling infrastructure, publisher relationships, ranking systems and a retrieval layer that can survive high query volume. Whoever controls that layer can become a recurring toll collector for training, inference and automated research. Keenable still has to prove that its claimed 100-billion-document index produces better answers in customers' workloads. Revenue, query volume and customer retention remain undisclosed. Buyers will need to test the service on their own workloads. Styskin and Petri have chosen a problem they know at production scale. The $26 million gives them enough capital to build beyond a thin wrapper around someone else's search engine. It also commits Keenable to the expensive version of the bet: owning the index, keeping it current and persuading AI developers that independence is worth adding another vendor to the stack.