You've been here before. You find a tool that would genuinely fix the thing that annoys you — your AI agent forgets everything between sessions, so you explain the same project, the same preferences, the same decisions every morning. You open the install guide and hit a wall: pip install, config files, something about "hooks", a warning that assumes you know what a hook is. You close the tab. Not because you're incapable, but because the tool was written for somebody else.
This is the pattern for nearly every developer tool on the internet: a wall of commands, and an unspoken assumption that a human types them. If you can't run terminal commands, the tool is effectively not for you — no matter how much you need it.
The assumption is wrong, and it's worth seeing why before we show you the exception.
Someone has to do the installing — it doesn't have to be you
Look at what that wall of commands actually asks of a person: fetch some code, put it where the agent can find it, connect it to the agent, run it once to check it worked. Every one of those is a thing your AI agent already does all day. It fetches things. It edits files. It follows instructions and reports back. Installing a tool is, frankly, an easier job than the ones you already give it.
The reason you'd ever type those commands yourself is history: most tools were built before agents, by developers, for developers, and nobody revisited the assumption. The natural order is that you describe the outcome — "install this memory tool" — and the agent does the work, the way it does everything else you ask.
That's not a hypothetical. It's how our memory tool is installed.
How it actually works: two human steps, one sentence
The tool is called brethof-brain. It's a memory system for AI agents — your agent remembers you across sessions, projects and machines. Here is the entire install, end to end.
Your part, first. Sign up on the site, choose your edition and copy your key from the panel (the hosted edition also asks you to set a passphrase). That's the whole account part — registering and clicking a button, which is what every application on earth asks of you. It is not IT knowledge.
Then your part, spoken. You open your agent and say:
Install brethof-brain.
It does the rest: it fetches the client, wires the hooks into your harness, and checks its own work. Once, a small window opens on your own screen where you paste the key (and, on hosted, your passphrase), so your agent never sees either. You never open a terminal. You never read a README. If something needs deciding, it asks you in plain language, the way it would for any other task.
We're specific about where the human steps end because it matters: they end at that window. There is no "and then it helps to have someone technical nearby." That caveat would be a lie dressed as modesty.
What you get the morning after
You start your PC, open a session, and say "continue." The agent already knows what you were doing — what you decided yesterday, what you prefer, where things stand. You don't re-explain. You just carry on.
This is also where the memory earns its keep over time. Your agent stops re-proposing the thing you rejected in March, because the line saying so — with the day and the reason — comes back with its search. When you ask "why is it like this?", the answer is in the words of the conversation that decided it. And if you change your mind, the reversal is a new line; the old way stays on record as the old way.
You also stop maintaining the pile of notes some people keep to remind their agent who they are — the folder of markdown files that grows until it's a jungle nobody can navigate. There are no files to tend. The memory keeps itself; every session opens with your agent already briefed. You don't manage anything. You talk, it remembers.
But I don't run servers — which edition, and what's on my machine?
Two editions, and you pick by comfort, not by skill:
Hosted. We run everything. No container, no disk space, no setup beyond the passphrase you set at signup. Nothing to run on your machine at all.
Local. Your memory lives on your own machine and nowhere else — we process it, we never store it. This does mean something runs on your computer: a single container, a self-contained box that holds your memory and its search. Your agent installs it, not you. You never configure it.
If "a container on my machine" sounds alarming and "we run it" sounds restful, pick hosted. If keeping the memory on your own disk matters to you, pick local. Both are installed by the same sentence.
Try it: the whole thing is one sentence
Sign up, choose your edition, copy your key. Then tell your agent:
Install brethof-brain.
Start free. If it works, you'll know by the second morning, when you say "continue" and it simply knows.
We built it this way because of a conviction we keep coming back to: AI should do the work for the human. Installing the tool that gives your AI its memory is work for the AI.
Disclosure: I build brethof-brain. How the install works, and the free plan (500 messages a month, no card): https://brethof.ai/brain/install/
If you ever want to recommend it, we run an affiliate programme through Lemon Squeezy: commission on every sale you bring, paid to you by Lemon Squeezy directly. Sign-up: https://brethofai.lemonsqueezy.com/affiliates
Top comments (0)