Engram

Engram is a San Francisco-based AI startup founded out of Stanford University's AI lab that builds a learned memory layer for enterprise AI systems. Instead of reprocessing organizational documents on every query, Engram pre-trains compact "Cartridge" memory objects on each organization's corpus — legal documents, codebases, Slack conversations, and policies — so AI agents can retrieve organizational knowledge at a fraction of the token cost. The company claims its models can match or outperform frontier systems using up to 100 times fewer tokens, and has early commercial partnerships with Microsoft (M365), Notion, and legal AI startup Harvey. Engram raised $98 million led by General Catalyst, Kleiner Perkins, and Sequoia Capital.