The development journey — from idea to cross-platform AI memory layer. Technical decisions, product evolution, and what's next.
Every AI tool you use starts from zero. We set out to build a shared memory layer that changes that — one brain for every AI.
The API worked, but nobody wants to make HTTP calls mid-conversation. So we built a Chrome extension that lives inside every AI chat.
Users wanted to see their memories, not just save and search. So we built a full web dashboard for browsing, organizing, and managing your AI memory.
Saving and searching memories is useful. But what if the system could learn your preferences, detect contradictions, and track your growth automatically?
A memory layer that only works in the browser isn't truly universal. We added MCP for IDEs and a CLI for terminals — and learned hard lessons about cross-platform identity.
A memory layer that stores personal knowledge needs serious security. We locked down the stack — rate limiting, encryption, GDPR compliance, and a full operational audit.
The product outgrew its original name. We rebranded to memset — a name that captures what we actually built: a memory layer you set once and it persists everywhere.
Unicode emojis out, Lucide icons in. Sharp corners out, rounded cards in. We overhauled every surface to match the new brand — and the product finally looked as good as it worked.
The missing piece: making IDEs proactively recall your memories and sync your preferences bidirectionally. We built it using MCP's instructions field and a new sync_preferences tool.
The foundation is built. Here's what's coming — a desktop system agent, mobile companion, and a public API that lets any developer build memory-aware AI applications.