Keenable

Search engines were designed around human attention spans: a list of links, a snippet, a click. AI agents can consume full documents and need fresh, structured, high-recall retrieval rather than a ranked list of ten blue links. Keenable is a startup building infrastructure for that use case: a massive web index whose API is already in production at several AI labs and inference providers. It came out of stealth in August 2026 with a $26 million seed round led by Accel.

The founders include Andrey Styskin, who previously led search, AI, and cloud at Yandex (Russia's dominant search engine). Their premise is that AI agents create a new retrieval flywheel, one shaped by how models process documents rather than by how humans click results. Keenable's index is consumed during both model training runs (to provide fresh data) and at inference time (so agents can look things up live as they execute tasks).

For builders, Keenable sits in the same layer as retrieval APIs and search-as-a-service tools, but is differentiated by being purpose-built for agent consumption rather than adapted from human-facing infrastructure. Think of it less as 'a search engine' and more as a document retrieval substrate that agents call the same way they would call any other tool.