Mirror

Mirror is a personal agent that learns how you think and choose, and lends that understanding to the AI tools you already use. The problem: every AI assistant starts from zero. You re-explain how you like documents written, what you reach for when you're tired, how you review code. And memory features that do exist give flat verdicts about you with no way to see why or correct them. Mirror takes what you tell it in your own words and keeps it as a time-aware knowledge graph. It works out what repeats and turns it into hypotheses, never flat statements. Each hypothesis carries the memories that support it, its counter-examples and its conditions. A pattern is only promoted after it holds on two distinct days with no contradiction. You can mark any hypothesis Correct, Wrong or Don't use, and deletes take effect immediately. You see it as a body map: the brain thinks and chooses, the hands write and make, the mouth eats. Tap a region to see what Mirror knows and the evidence behind it. Other AI tools get only the relevant part, through a prompt hook (Claude Code, Muse Code), MCP, a plain HTTP API or an OpenAI-compatible gateway. Sensitive patterns (relationships, self-understanding) are marked internal-only and never leave Mirror. Each user runs in their own container with their own data, so users never see each other's memories.