Keenable aims to build a search index optimized for AI agents with $26M seed round

The web's dominant search infrastructure was purpose-built for human users—people who lack the capacity to wade through entire pages. As artificial intelligence systems increasingly perform research and retrieval tasks, a growing wave of technologists believe the internet's foundational indexing and search layer needs reimagining to serve machines instead. AI models can absorb far more text and data than any person could, creating an opportunity for search systems designed with those capabilities in mind.

That's the wager behind Keenable, a new startup co-founded by Andrey Styskin—former head of search, AI, and cloud operations at Russian search powerhouse Yandex—and Matthias Petri, an AI researcher from Germany. The pair recently emerged from stealth with $26 million in seed funding, led by Accel with backing from Conviction Partners and several individual investors.

The case for AI-native search#

Styskin sees a fundamental advantage when AI systems can retrieve and cite source documents to support their outputs. As he explained, "This actually creates a new flywheel that is different from what Google learned from human behavior." Rather than optimizing for snippets and summaries, an AI-optimized index can expose much richer context, helping language models ground answers in evidence.

Keenable has already assembled a massive corpus: over 100 billion documents indexed. The API is active in production across several AI labs and inference providers, both during model training and at inference time, though the company has declined to name specific customers. The startup did recently announce a partnership with voice AI firm Gradium to enable live information retrieval for conversational agents.

The infrastructure gap#

Major tech companies are stepping back from their own search APIs. Google has shut down Custom Search in certain configurations, while Microsoft has retired its Bing Search API, both moves aimed at protecting their core search franchises. These choices leave AI developers with few options for large-scale web search.

Styskin's background—20 years building search infrastructure at Yandex and Amazon—gives him insight into the engineering challenges. The naive approach of indexing and scanning the entire internet at scale becomes prohibitively expensive without careful optimization. "If you do not fine-tune your index structures for a specific task, the cost of serving and scanning the whole internet is enormous because of the volume," he said. Keenable's edge lies in rapidly narrowing the search space based on what a query actually needs.