Mneme

End-to-end AI memory system solving context loss and invisible retrieval

Mneme is a custom interface and memory system for Anthropic's Claude. Designed and built over five months, it runs locally, handles context retrieval automatically, and gives the AI a layered, dynamic, visible brain.

Every conversation picks up where the last one left off and the UI clearly shows what's happening.

Challenges

Most AI chat interfaces are built for casual use. Conversations reset, the context window runs out, and even when memory features exist, they aren't visible. As a user you don't know what the system is doing, or if it's doing anything at all. For people who use LLMs often, losing continuity is really the only thing that's continuous.

Process

I wanted to build something that felt native to the Claude ecosystem while establishing its own identity.

The logo's geometric folds represent information density, the folds of a brain, a sine wave and a single continuous line. The processing animation uses the same logic with a circuit diagram look. Idle is a flat line. Processing turns it into crawling folds. Tapping the circuit breaker button flattens the wave and visually disconnects the system. Three logo variants' fold density scales with model complexity.

The backend layers five main types of memory: semantic search for past messages, entity summaries to track people and topics, keyword-triggered narrative concepts, daily rolling summaries, and persistent notes the AI maintains itself.

I wanted these mechanics to be transparent, solving the UX problem of waiting and hoping your memory system works. I built visibility directly into the interface so you can watch what retrieves, when entities trigger, and how tokens distribute, in real-time. Watching the context panel during the processing wait teaches you how the memory system works. You don't need documentation to understand it.

(The scripted demo at the bottom of the page is interactive.)

Scripted demo