Engram is an AI company developing models that learn from the information and context people use in their work. Rather than relying only on general training data or repeatedly searching through documents, Engram aims to help models absorb knowledge from large, shared workspaces and improve as that context changes. Its approach starts with pre-trained models and applies training compute to the data that matters to each organization, with the goal of making AI more informed, efficient, and useful over time.

Engram’s first product is an API for agents designed to learn from large knowledge workspaces. The company is working with Notion on custom agents that understand extensive Notion workspaces, with Harvey on models that internalize firm knowledge and identify precedents across client matters, and with Microsoft on a pilot embedding Engram models in Microsoft 365 for customized, cost-efficient enterprise agents. Engram says its models can connect information across GitHub, Slack, and Notion, helping agents retain relevant knowledge and avoid repeatedly gathering the same context.