Memex runs as a local desktop application rather than a browser-based or cloud tool, which is its core differentiator from most AI app builders: projects and data stay entirely on the user's machine, and every AI action can require explicit approval or run fully autonomously, depending on how much trust the user wants to extend.
It offers a Chat mode for planning and discussion and a separate Build mode for actually executing changes, supporting essentially any tech stack rather than locking users into one framework, with Python, React, AWS, and various cloud databases all fair game. Template-based scaffolding and project-specific rules give it some of the structure of a framework while keeping the local-first, privacy-focused approach as the main selling point.
The company lists organizations including Google, Harvard, Salesforce, and UCLA among its users, suggesting adoption beyond hobbyist projects, though as with most private usage claims from vendors, the specifics of how deeply embedded it is at those organizations aren't independently verifiable.