EN
Back to the archive

The archive · AI & Models · Strategic decision · 2026

Keenable bets AI agents need their own web index; $26M seed from Accel

Ex-Yandex search chief Styskin builds a 100B-document index for AI agents; Accel leads $26M seed as Google and Microsoft close search APIs.

Keenable

The betThat AI agents need search built for machines, not humans — an independent, cheaper index can beat Google on agentic queries where it never could for human search.Live

What the business is

Keenable builds a web-scale search index and retrieval API designed for AI agents, so models and inference providers can ground their answers in live source documents.

Starting capital$26M seed (Aug 2026) led by Accel with Conviction Partners and angels from Amazon, ClickHouse, Databricks, Google, Snowflake, SpaceXAI

How it started

After 20 years building search at Yandex — where Styskin rose to CEO of Search and built a 200-billion-document index — and at Amazon AGI, where he and Petri built web search for Alexa, the pair saw AI crawlers taking a growing share of search volume. They concluded search had to be rebuilt for machines, not humans, and founded Keenable.

What happened

The company emerged from stealth on 2026-08-25 with a $26M seed led by Accel and participation from Conviction Partners. It says its index covers more than 100 billion documents, its API is already used in production at several AI labs and inference providers during training and runtime, and it has partnered with voice-AI company Gradium for live information retrieval. Keenable is also building Web Query Language, a product that lets AI systems combine answers across web sources. With 15 engineers across the US and Europe, it plans to double headcount by the end of 2026.

How it ended up

Still live and early: Keenable plans to double its 15-person team by end of 2026 to build go-to-market; it has not disclosed revenue or named customers.

Background

Keenable is a search-infrastructure startup founded by Andrey Styskin, former CEO of Yandex Search, and Matthias Petri, formerly a principal applied scientist at Amazon AGI. It emerged from stealth on 2026-08-25 with a $26 million seed round led by Accel, with participation from Conviction Partners and angels from Amazon, ClickHouse, Databricks, Google, Snowflake and SpaceXAI.

The company's bet is that AI agents need an index built for machines, not people: chatbots answer better when grounded in source documents, and they query at far higher frequency and lower latency than human searchers. Keenable says it has built an independent index of more than 100 billion documents, with an API already in production at several AI labs and inference providers during both training and runtime.

Styskin argues Google is 'beatable' on agentic queries because the innovators' dilemma keeps it from serving AI customers as aggressively as an independent player can. Keenable is building Web Query Language so agents can combine answers across thousands of live sources, and has partnered with voice-AI company Gradium. With 15 engineers across the US and Europe, it plans to double headcount by the end of 2026.

What has to be true

  • AI agents retrieve differently from humans — high-frequency, low-latency, source-grounded — so a human-first index is structurally mismatched to them.
  • Google and Microsoft shutting down search APIs to avoid cannibalization left AI companies few independent infrastructure options.
  • The founders already built search at web scale: Styskin ran Yandex's 200-billion-document index and beat Google twice; at Amazon AGI they built search for Alexa.
  • Several AI labs using the API in production before the round closed was evidence of demand, not just narrative.

What can be applied

When incumbents close the APIs everyone depended on, the opening is to rebuild the layer from scratch for the new customer — incumbents can't serve a new workload without cannibalizing the old one.

Aftermath

As of 2026-09-02 Keenable is live and growing: a 100B+ document index, API contracts with multiple AI labs, a Gradium partnership, and $26M to double its 15-person engineering team by the end of 2026. It has not disclosed revenue or named customers; its stated ambition is to become 'the next Google for AI agents,' competing with Brave and Exa as Google itself overhauls search for the AI era.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases