What is personal-model?

Persome, also known as personal-model, is a tool that builds what it calls your HUMAN.md: a private model of how you actually think and work, built from your own activity on a Mac. It watches focused activity you allow it to see, after you grant macOS permission, and turns that into a structured picture of you that AI tools like Codex and Claude Code can read through MCP, the Model Context Protocol. The project describes this model as growing through layers. A Point is a single sourced observation. A Line connects related points into a relationship or change over time. A Face is a pattern backed by several related points. A Volume is a larger structure spanning multiple projects or areas of life. The Root is the current combined model of you, built from all of that evidence. Every claim in the model keeps a link back to the observation it came from, so nothing is asserted without a source, and new evidence can update or overturn earlier conclusions. The stated goal is that the same AI assistant, given the same prompt, should respond differently depending on who is asking, because it understands that person's goals, priorities, working habits, and boundaries. Connected coding agents can use this model to find unfinished work, prioritize it, and separate tasks that are safe to do locally from decisions that need a human to weigh in. Persome runs entirely on your Mac and is described as private by default, with your data yours to inspect, correct, export, or delete. It requires macOS 13 or newer. There is a five minute synthetic demo you can try without touching any of your real data, using a sample script that opens a local viewer and then deletes the demo data afterward. Installing with your own data uses a command line tool called persome, and a large language model is optional for basic use but required if you want the deeper semantic modeling features. The code is released under the Apache-2.0 license.