Start here

Neuronz.ai gives your AI a memory that lasts beyond one conversation: what your projects are, how you like things done, what was decided and what was already tried. This section is a short course. First you see what you can get and where, then you set it up, then a few tutorials show you what to do with it.

You need a Neuronz.ai account. Create one — it takes a minute — then come back here.

The model and the harness

Two things are at work when you talk to an AI:

The modelThe AI itself — Claude, ChatGPT, Gemini, or an open model you run yourself. It reads and writes the text.

The harnessThe app you type into — claude.ai, the Claude desktop app, Claude Code, Oh-My-Pi. It sends your messages to a model and gives it tools.

Neuronz.aiPlugs into the harness, so whichever model it runs can read and write your memory.

Your memory lives in Neuronz.ai, not in the model, so it stays the same whichever model you use. What changes from one harness to another is which models you can pick and what Neuronz.ai can do in it.

Models and features differ by harness

Which models you can use. Some harnesses, like Oh-My-Pi, work with almost any model out of the box. Others are built for one maker's models but can be set up to reach others, as Claude Code and Codex can. And some are locked to their own: the Claude, ChatGPT and Gemini apps only ever run their maker's models.

What Neuronz.ai can do there. Some harnesses work with Neuronz.ai and some do not yet, and the ones that do don't all offer the same things. In a chat, the memory is there when you ask for it. In a coding harness like Claude Code, it works on its own: it checks your memory before each message, follows your rules and saves what happened. Each page under Connect your agent says what you get in that harness.

Compared with the app's own memory

Most harnesses already remember something. Claude and ChatGPT keep a memory of your chats, and Claude Code keeps notes and reads a CLAUDE.md file from your repository. These memories are mostly one block of text that grows as you use them, and the harness loads it whole at the start of each chat or session, whatever you're about to do. The more it holds, the more of every task goes to reading notes that have nothing to do with it.

Neuronz.ai works differently:

One thing per recordA fact, a rule, a decision or a document is saved on its own, not appended to a file that keeps growing.

Only what the task needsThe agent searches your memory and reads only the records that match what you're doing, so a large memory doesn't weigh on a small question.

Kept up to dateWhen something changes, the new record replaces the old one instead of both sitting side by side, and contradictions are flagged for you to settle.

One profile, many foldersA harness ties its notes, rules and skills to one folder. A profile can cover several folders and brings your records, rules, skills and commands into each of them.

You don't have to turn the app's own memory off: keep a short CLAUDE.md for what every session in a repository must read, and let Neuronz.ai hold the rest.

Connect your agent

Pick the harness you will use first and follow its page. You can add the others later with the same account.

Claude on the webChats at claude.aiClaude desktop appIts chats, and its Code tabClaude mobile appChats on your phoneClaude CodeIn a terminalOh-My-PiIn a terminal, with almost any model

Then the tutorials

Once it is set up, each tutorial takes a few minutes and works in any harness — where a step differs, it says so.

1. Split your memory into profiles — work, personal, finance: what a profile is and how your harness picks one. Start here; everything else lives in a profile.
2. Remember once, pick up later — tell it something once, find it again next week.
3. Keep a piece of work in one topic — give an ongoing piece of work its own place, and pick it up the next day from its brief.
4. A rule for one situation — something the agent must always do, but only when it applies.
5. Your voice, then a message — teach it how you write, then have it send a Slack or WhatsApp message for you.
6. Learn a code repository — fill a profile from a codebase, and how that differs from Claude Code's own /init.
7. Learn your business from your company wiki — the same, for what your business knows instead of code.