Codex memory that follows you to the next machine

Codex has a memory now. Its Memories feature, off by default, turns earlier chats into notes it reads in later ones, and one store serves every project on the machine. OpenAI’s docs say where that store lives: "Codex stores memories under your Codex home directory. By default, that's ~/.codex."

Its own Memories stay on the machine that wrote them, so connect Codex to Pando and have it write decisions into your outline, which every machine reaches, and so does every other agent you allow to read it.

Set it up

The Codex CLI, its IDE extension and the ChatGPT desktop app share this setup, so one add is enough on each machine.

What each place is for

• AGENTS.md, for rules the whole team must follow. OpenAI’s own advice: "Keep required team guidance in AGENTS.md or checked-in documentation."
• Codex Memories, for recall on one machine. The docs call those files "generated state" and say not to rely on editing them by hand.
• A branch of your outline, for decisions and their reasons that have to survive a second machine, a second agent, or a question you ask from your phone.

Why a memory you can read

A memory you cannot correct is one you learn to distrust. In Pando each entry is a bullet. You read it on your phone, fix the line that went wrong, and the next session reads the fixed one.

Working with other agents

Claude Code, Claude and ChatGPT connect to the same outline. What Codex decided on Tuesday is what Claude Code reads on Friday, which a store under one machine’s home directory cannot do.

What it costs

Nothing, for 1,000 bullets, with agents included and no card. The tokens are paid by your own OpenAI plan, so Pando has nothing to resell and no reason to meter the calls.

Deeper

• Claude Code memory on every machine you use
• Taking the memory with you

Connect an agent, about two minutes · The twelve tools · Which note apps an agent can reach · Who runs this