Open-source desktop agent app and Rust library for running many AI agents with lower token use, memory and resource cost.
Intuition-Lab/personal-model
Local-first, macOS long-term memory runtime that builds an evidence-linked personal model from your activity and serves it to Claude Code, Codex, and other MCP clients.
- Stars
- 1,337
- Forks
- 91
- Open issues
- 15
- Last commit
- 6 days ago
What is it?
Personal Model captures focused activity on your Mac after you grant macOS permissions, then organizes it into layered context: sourced Points, relationships (Lines), recurring patterns (Faces), higher-order structures (Volumes), and a current integrated model called the Root. The result is exported as a HUMAN.md-style document that AI tools can read.
The model runs on your machine and is exposed through MCP, so trusted clients such as Claude Code, Codex, Cursor Agent, Claude Desktop, and opencode can search it, open source receipts, and ground their answers in your own history. Claims keep provenance, corrections preserve audit history, and you can inspect, correct, export, or delete the data.
LLM access is needed for semantic modeling and can be configured with a hosted or local provider, or by lending an existing coding-agent subscription with a daily call cap. Collection and BM25 recall work without one. Installation uses the uv package manager or an install script and requires macOS 13 or newer.
Key features
- 01Captures focused activity on macOS after explicit permission grants.
- 02Builds layered models from Points, Lines, Faces, Volumes, and a Root.
- 03Exposes memory to Claude Code, Codex, Cursor Agent, and other MCP clients.
- 04Keeps source receipts so claims can be traced back to evidence.
- 05Lets you inspect, correct, export, and delete your data.
- 06Runs locally, with optional hosted or local LLM providers for semantic modeling.
Use cases
- Giving Claude Code or Codex context about your current projects and priorities.
- Asking an agent to find unfinished work and propose next steps grounded in your history.
- Keeping one personal memory shared across several MCP-compatible coding tools.
- Reviewing and correcting what a personal AI has inferred about your working patterns.
Quick start
- 1
Install the package with uv
uv tool install personal-model - 2
Run interactive onboarding to grant macOS permissions and set up the local Runtime
persome onboard - 3
Open your Personal Model after 30 minutes of activity
persome model open --after 30
Things to know
- Runs only on macOS 13 or newer, and requires Xcode Command Line Tools.
- Semantic modeling requires a configured LLM provider or a coding-agent subscription, which may carry cost or a daily call limit.
- It needs macOS Accessibility and, for some capture modes, Screen Recording permissions.
Topics
FAQ
- What is personal-model?
- Local-first, macOS long-term memory runtime that builds an evidence-linked personal model from your activity and serves it to Claude Code, Codex, and other MCP clients. Personal Model captures focused activity on your Mac after you grant macOS permissions, then organizes it into layered context: sourced Points, relationships (Lines), recurring patterns (Faces), higher-order structures (Volumes), and a current integrated model called the Root. The result is exported as a HUMAN.md-style document that AI tools can read.
- Is personal-model free to use? What license does it have?
- personal-model is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is personal-model still maintained?
- Actively maintained. The last commit to personal-model was on Oct 16, 2026.
- What can I use personal-model for?
- Giving Claude Code or Codex context about your current projects and priorities. Asking an agent to find unfinished work and propose next steps grounded in your history. Keeping one personal memory shared across several MCP-compatible coding tools. Reviewing and correcting what a personal AI has inferred about your working patterns.
- How do I install personal-model?
- Install the package with uv: uv tool install personal-model Run interactive onboarding to grant macOS permissions and set up the local Runtime: persome onboard Open your Personal Model after 30 minutes of activity: persome model open --after 30
- Where can I get personal-model?
- The source code and installation instructions are on GitHub: https://github.com/Intuition-Lab/personal-model
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A local-first AI companion daemon for Windows and macOS that keeps long-term memory of you, turns chats into office files, and can extend itself.
Self-hosted private 1:1 chat channel linking a phone PWA to your AI companion, running via Claude Code or any OpenAI-compatible LLM.
A local-first, provider-neutral personal AI workspace with chat, reviewed memory, auditable tools, MCP support, and isolated project plugins.
A self-hosted personal AI agent where each agent runs in its own Linux desktop container on your computer, with any model, an encrypted vault and Telegram control.
Self-hosted long-term memory service that gives persistent AI identities continuity across clients, models, and agent interfaces.