Personal Model
A local-first, long-term memory layer for coding agents that learns how a developer works from activity captured on their Mac and shares that context across Claude Code, Codex and Cursor.
Personal Model builds an evidence-linked profile of how a person thinks and works, called HUMAN.md, from focused activity it captures on a Mac after the user grants macOS permission. It gives connected AI tools inspectable context to continue work and make grounded decisions, rather than relying only on chat history.
The model organizes what it learns into five layers of increasing scope: a sourced observation, a relationship or change over time, a pattern supported by related evidence, a higher-order structure across projects, and the current integrated model of the person. Important claims keep their source evidence, so new evidence can strengthen, revise or overturn an earlier inference instead of silently replacing it.
It connects to Claude Code, Codex, Cursor Agent and other MCP-compatible clients through native installers or a managed stdio configuration, and it runs locally, keeping the model private and letting the user inspect, correct, export or delete it without depending on a hosted memory service. It is written in Python and released under the Apache-2.0 license.