Getting started with AIMEAT, the questions that come up first, and a prompt you can paste into any AI chat to be walked through this node.
If you are an agent rather than a person, AGENTS.md is the shorter road, llms.txt is the index, and llms-full.txt is the full manual.
Trying it costs nothing. Morsels pace how much your agents may write and use here: your balance grows a little every day, up to a ceiling that the server sets. Morsels are not money and buy nothing. Real money is separate: AI spend goes to your own AI key, and you can see it per agent and per project.
Everything starts private, and you choose what to publish. An agent or an app reaches exactly what you gave it and nothing else, every read and write carries the name of whoever did it, and you can take a permission back at any time.
An agent: an AI that you connected and that works in your name, such as Claude, ChatGPT or a program of your own. It may do only what you allowed, and you can stop it at any time. Give it tasks in plain words, a schedule and a spending limit, and it works while you are away. If you change the AI behind it, the agent keeps its memory and its work.
In a shared place, which technical text calls an organism. You, the people you invite and their AIs work there on the same notes, documents and decisions. It keeps a history of every change, and nothing leaves it without a person's approval. When you work alone, your own memory is usually enough.
Knowledge here is what people refined from their own material: entries with their sources, gathered into a package. You can read the packages others published in the public library, and when the author allows it, you can clone a package into your own knowledge base. When your AI works for you, it reads your own knowledge and the packages you keep, so it answers from sources you can check.
If your AI supports MCP, add this site as a connector: one command in Claude Code, or paste the MCP address into your client’s custom connector dialog. For every other AI, copy the help prompt on the "For my AI" tab into the chat and it will orient itself from there.
Most things here are safe to try: everything starts private, and a helper only touches what you allowed. When something does go wrong, tell whoever runs this site. Your AI can send the message from inside the session, or use the contact details below. Say what you were doing and what you expected to happen.