Skip to main content

A Foundation for Building Personal AI Systems

Comparison: Traditional architectures where the App is the platform versus this architecture where Your Memory is the platform
Learn How It Works

Four Components

Each one is independent. Each one is swappable. Your Memory is the only one that stays.

Your Memory

The platform. Your data, config, preferences, and conversations. Everything else depends on it — it depends on nothing.

Agent Loop

The agent loop. Sends messages to the model, executes tool calls, repeats until done. New capabilities arrive by adding tools.

Gateway

The front door. Web app, CLI, mobile app, Discord bot — any interface connects through it. Start on one, continue on another.

Auth

Controls access to your system. Independent of every other component — swap anything else, auth doesn't change.

Two APIs

Gateway API

How your interfaces connect to the system. Any client that speaks the protocol works — web app, CLI, mobile app, Discord bot.

Model API

How the agent loop connects to AI models. Swap providers with a config change — no code changes required.

Try Building On It

We're building a product on this architecture ourselves and sharing what we learn along the way. We're looking for a few developers who care about this problem enough to do the same.

1

Read the Architecture

Understand the four components, two APIs, and why every piece is swappable.

Foundation Spec
2

Build a System

Use the reference implementation as a starting point.

npm install @personalaiarchitecture/core
3

Share What You Find

What worked, what broke, what's missing — we want to hear it.

Join the Forum