myAI · animated tour

How myAI's brain works

Every AI session normally starts from zero. myAI gives every session — Claude, Codex, Antigravity or a local model — the same memory, the same tools and the same to-do list. Here is the whole loop in under a minute.

Scene by scene

The same eight scenes, one at a time — best for reading on a phone.

1. Scene 1

   08:00 — every Mac wakes itself up

   At eight every morning each machine starts its own copy of myAI, pulls the latest code and the latest memory, checks that everything is healthy, and writes a one-page report. By the time I sit down, a session is already open and waiting.

2. Scene 2

   Memory that works like git

   The brain is a git repository. Every working session writes its notes as a commit; "wrap up" merges them in. Nothing is overwritten, every memory has a history, and any machine can rebuild the whole brain from the remote copy.

3. Scene 3

   Distilled, so it stays small

   Thousands of notes would drown a new session, so the brain distils them: a one-line brief, the active working context, and a longer roll-up. A fresh session reads the short version first and only digs deeper when it needs to.

4. Scene 4

   "Have we solved this before?"

   Past sessions are stored as searchable patterns. Before an agent starts writing code, it asks the brain whether this problem has come up before — and gets the earlier fix back instead of rediscovering it.

5. Scene 5

   One gateway for every AI tool

   Every AI tool connects through one small local gateway that speaks the open MCP standard. Through it they all share the same task queue, the same memory, a mailbox to message each other, and the swarm — and nothing gets in without its token.

6. Scene 6

   Work happens while I'm away

   A runner on my own machine works through the queue after hours and leaves finished changes for me to review. The checks run on my own self-hosted CI runner too, instead of paying for cloud minutes.

7. Scene 7

   One goal, many models

   A bigger goal becomes a swarm: up to five smaller tasks, run in order, spread across Claude, Codex, Antigravity and local models. Nothing starts until I approve it, it has a spending budget, and one button stops all of it.

8. Scene 8

   Tomorrow starts smarter

   Every wrap-up writes down what was learned. The next morning's boot starts from there — so each day begins a little further along than the last.

Play with it

Why it matters

The point isn't any single feature. It's that the work compounds: memory survives the session, the machines keep themselves ready, and the same context reaches whichever model is best for the job. More about how it's built in myAI, in depth.