Keenable Secures $26M to Build Search Index for AI Agents

Keenable, a new startup founded by industry veterans Andrey Styskin and Matthias Petri, has officially emerged from stealth mode with a significant $26 million seed funding round. The investment was led by Accel, with participation from Conviction Partners and various angel investors. The company is focused on a specific, high-stakes problem in the current AI landscape: providing a reliable, cost-effective, and performant web search index tailored specifically for autonomous AI agents. Unlike traditional search engines designed for human consumption, Keenable is architecting its infrastructure to support the unique retrieval patterns required by large language models during both training and inference phases.

Styskin, who previously led search and AI divisions at Yandex, and Petri, a German AI scientist, are leveraging their deep experience in search infrastructure to address the limitations of current enterprise solutions. The core technical challenge they are tackling is the prohibitive cost and latency associated with scanning the entire internet for every query. Traditional enterprise search tools often struggle to scale effectively, leading to high operational expenses. Keenable claims to have already indexed over 100 billion documents, offering an API that is currently being utilized by several AI labs and inference providers. A notable early partnership includes a collaboration with Gradium, a voice AI company, to facilitate real-time information retrieval.

This development comes at a time when the search landscape is shifting. As AI models become more agentic, they require constant access to fresh, grounded data to minimize hallucinations and improve accuracy. However, major incumbents like Google and Microsoft have become increasingly restrictive with their search APIs, often limiting access to prevent cannibalization of their own consumer-facing products. This creates a vacuum for specialized infrastructure providers. According to Accel partner Zhenya Loginov, the lack of open, web-scale search infrastructure for AI developers is a critical bottleneck that Keenable aims to resolve.

Keenable is also developing a proprietary technology called Web Query Language. This tool is designed to help AI systems synthesize information across multiple web sources, even when no single document contains a complete answer to a complex prompt. This represents a departure from the standard retrieval-augmented generation (RAG) workflows that often rely on simple keyword matching or basic vector similarity. By narrowing the search space intelligently, Keenable hopes to offer a more efficient alternative to the brute-force methods currently employed by many developers.

Despite the technical promise, the company faces significant hurdles. Building and maintaining a search index of this magnitude is notoriously expensive and resource-intensive. Styskin has acknowledged the financial strain of such an endeavor, noting that the company is carefully managing its capital to scale its engineering team, which currently stands at 15 people across the United States and Europe. The startup plans to double its headcount by the end of the year to accelerate its go-to-market strategy.

Competition in the space is intensifying. Other players, such as Brave and Exa, are also vying for dominance in the AI-native search market. Furthermore, Google is actively re-engineering its own search experience to accommodate the needs of AI agents. The success of Keenable will depend on its ability to prove that its specialized index offers superior performance and lower costs compared to the massive, general-purpose engines controlled by tech giants. As the industry moves away from the traditional ten blue links model, the race to define the infrastructure for the next generation of AI-driven information retrieval is clearly underway.