Keenable

Keenable is an AI infrastructure company that provides web search and knowledge access for AI labs, inference platforms, and agent builders. It offers a Search API exposed over MCP, a SELECT product that turns web pages into queryable tables, and a Time Machine point-in-time search API. The homepage states the index covers 100B+ documents with p95 latency under 250ms (US East).

What It Is

Keenable is web search infrastructure designed for AI agents rather than human "10 blue links" results. Its About page says the company is building an independent web index and researching new access paradigms so agents can query the web cheaply and frequently. Agents connect through an MCP endpoint that exposes search and page-fetch tools, and the company lists cloud and on-premises options for larger deployments.

Three Products on One Index

• Search API: MCP tools named search_web_pages and fetch_page_content let agents search and retrieve page content.
• SELECT: A Web Query Language with functions such as WEB_SEARCH, SEM_EXTRACT, SEM_MATCH and SEM_NORM. It builds a table of values from many web pages, with each row citing its source URL.
• Time Machine: A point-in-time search API where adding query_time rewinds both the corpus and the ranking, allowing searches across prior versions of pages. It is offered through early-access requests.

Quality Claims and Integrations

Keenable publishes a chart comparing result quality and cost across search providers using its NEEDLE benchmark, which is a vendor-reported comparison. Its blog announces a web-fetch backend integration for cognee agent memory and a partnership with Baseten to make web search native to inference.