Universal Memory Protocol – a shared format for agent memory

A new project called Universal Memory Protocol is pitching itself as the missing link for AI assistants: a shared way for them to remember things when you switch apps, tools, or services. In plain English, the idea is simple and appealing: if your AI helper learns your preferences in one place, that memory shouldn’t vanish the second you move to another. The creator says the goal is user-owned, portable memory instead of today’s messy pile of notes, files, databases, and locked-in systems.

But in the comments? Absolute food fight. Supporters see a future where your AI finally stops acting like it has amnesia every time you change software. Critics, though, came in swinging. One of the harshest replies blasted the whole thing as “LLM slop rather than engineering work,” arguing the site has big promises but not enough proof, charts, or hard results. Another commenter openly laughed at the project’s claim that memory can be made “injection-resistant,” basically saying, uh, that’s not how these systems behave in real life.

And then came the classic standards-war energy. One person argued this whole protocol is overkill and that a simple shared folder layout would do the job better. Another dropped the legendary xkcd standards comic, which is internet shorthand for: congrats, you may have invented one more standard to compete with all the others. So yes, the dream is persistent AI memory — but the real spectacle is the comment section asking whether this is the future, or just another shiny tech acronym with main-character energy.