AI agents are pushing the limits of what traditional search engines were designed for. Recognizing that chatbots require vastly more comprehensible input than humans skim, Keenable is building a new kind of web infrastructure optimized for AI-equipped agents. The startup, founded by former Yandex and German AI researchers, has Surface-launched with a $26 million seed round led by Accel. Conviction Partners and angel investors also joined in.
A web index made for AI
Keenable says it has indexed over 100 billion web documents. Its API is already powering production environments at several AI labs and inference platforms. The firm recently teamed up with voice-AI developer Gradium to enable agents to pull in real-time data from the web.
Co-founder Andrey Styskin, who led search, AI, and cloud efforts at Yandex, and AI scientist Matthias Petri are positioning Keenable as distinct from enterprise search tools. The key, Styskin argues, is structuring an index that zeroes in on relevant content instead of blindly scanning the web. He emphasizes that cost efficiency is critical for serving live agent-level search at scale.
Opportunity in the search ecosystem
With Big Tech players scaling back access to search APIs and embedding proprietary search services into closed bundles, the need for open, scalable infrastructure is growing. Accel’s investment partner noted that AI developers currently have few alternatives if Microsoft and Google limit external access to web search data.
Keenable is also working on a feature called WebQueryLanguage, which will let agents merge information from diverse sites to answer questions when no single page has the full picture. The startup’s engineering team operates between Europe and the U.S., numbering roughly 15 people now, with plans to double by year’s end.
Building a full web-scale search index comes with massive costs. Styskin acknowledges the expense but believes Keenable’s innovation might allow it to compete in what he describes as “agentic queries”—the kind of searches AI agents handle. He sees opportunity to outperform incumbents on cost and specialized capabilities.
Other startups like Brave and Exa are also stepping into this space. Even Google is rethinking search for the AI era. Keenable believes the familiar format of “ten blue links” may soon give way to agent-oriented retrieval models.
Why this matters: As AI agents become more central to user interactions, the infrastructure behind them needs to change. Keenable’s effort reflects a broader shift away from search optimized for human eyeballs towards solutions tailored for machines. The next few years will determine who dominates AI-driven access to web information—and whether open models can rival the closed systems of major platforms.
