Keenable has emerged from stealth with $26 million in seed funding led by Accel and a plan to build web-scale search infrastructure for AI agents.
TechCrunch reports that the startup was founded by Andrey Styskin, previously a leader in Yandex’s search, AI, and cloud division, and German AI scientist Matthias Petri. Conviction Partners and angel investors also participated in the round.
The company says it has built a web-search index of more than 100 billion documents. It also says its API is already used in production by several AI labs and inference providers during training and runtime, though the report does not name those customers.
Agents need retrieval built for machines
Keenable’s pitch starts from a simple distinction: human search and agent search are not the same workload. Human search engines are optimized around people scanning a small set of links and snippets. Agents can read and synthesize much larger document sets, but they need retrieval systems that can narrow the search space quickly and cheaply.
That makes the index itself infrastructure. For an AI agent, retrieval quality can determine whether it grounds a response in current evidence, misses a key source, cites weak material, or spends too much time and money scanning irrelevant pages.
The startup is also building proprietary retrieval capabilities, including a planned Web Query Language product for answering questions that require combining information across multiple web sources.
The market timing is part of the story. TechCrunch notes that Google and Microsoft have been restricting or shutting down some existing search API paths while bundling more search capability into their own AI products. That leaves AI labs, inference providers, and agent builders looking for alternatives that are designed for machine-scale retrieval.
The strongest claims are still company-reported
The report gives enough to establish Keenable as a real infrastructure startup: funding, founders, stated product direction, a large index-size claim, and at least one named partnership with voice AI company Gradium.
But the customer and performance claims remain thin. Keenable does not disclose which AI labs or inference providers use the API, what latency or freshness guarantees it offers, how it handles publisher access rules, or how its retrieval quality compares with Exa, Brave, Google, Microsoft, and internal lab crawlers.
That does not make the story weak. It makes the right takeaway narrower: agent retrieval is becoming a distinct infrastructure market, and investors are funding teams that believe the search layer should be rebuilt around AI workloads.
