Agentic web search infrastructure startup Keenable raises $26M

Keenable.ai is opening the doors with a simple bet: the web search stack built for humans is already wrong for machines. The startup says it has raised $26 million and is using that money to push an index designed for autonomous agents that fire off lots of fast, narrow queries instead of the slower browsing patterns people use.

The company’s pitch rests on scale and speed. Its independent index covers more than 100 billion documents, and it offers APIs aimed at use cases like market mapping, pricing monitoring and lead enrichment. Those tools accept natural-language queries, fetch cleaned content and are meant to return “model-ready inputs” with minimal delay. Keenable says it prices the service at $1 per 1,000 API requests for “frontier-scale” users.

There’s also a historical angle. Keenable supports point-in-time queries, so a model can search the web as it looked on a specific date instead of only seeing the present tense version of the internet. And the company is building a product called Web Query Language to help AI systems answer questions by combining pieces from multiple sources when no single page has the full answer.

The founders come with search pedigree. Chief executive Andrey Styskin previously worked at Yandex, and chief scientist Matthias Petri came from Amazon AGI, where he focused on internet retrieval systems. Styskin said he saw the need for this kind of infrastructure while working on search for AI applications at Amazon, including Alexa. He argues that if an index is not tuned for a specific task, scanning the whole web gets expensive fast.

Keenable’s investors include Accel, Brightwing Capital, Conviction Partners and scOp Venture Capital, with angel participation from unnamed Google and Amazon executives. The company says some AI labs are already using its APIs in production, though it won’t name them. It has 15 employees now and plans to double that as it chases a market where many agents are still leaning on old search APIs or homegrown crawlers.