Your mail becomes tasks with dates. The work the desk proposes arrives already drafted, and waits for your yes. You press Run, and the agent takes it under your own login.
The desk reads your mailbox — Gmail, Outlook or any IMAP account — and writes what it finds into one queue. A thread that needs an answer becomes a task with a date on it, not a flag you scroll past.
The work the desk proposes arrives already written: the reply, the offer, the post. It sits there and waits. You approve it, you edit it, or you send it back.
The agent starts on your machine, under your own login, and stops when the task ends. The task says whether it is running, waiting on you, stopped or done. Nothing starts on its own.
Your pipeline, your outreach, your invoices, your money in and out, your tickets, the customers you are onboarding and the content you owe are pages on the same desk. Each one carries the count of what is waiting for you. Each one downloads as a spreadsheet, so your accountant never needs a login.
A task holds the tasks under it, so one record carries a whole job. A plan you post from your own terminal, with your own key, becomes tasks with dates on the same chart.
Drag a task above the one beside it, or drag a card into another column. The desk keeps the order you left it in, and a task you drag under another becomes part of it.
A task that waits for another draws a line to it. The chain that decides your end date is drawn heavier than the rest, so you can see which task to move first. A date with no work behind it is a diamond on the same chart.
A run launches the agent with every permission granted. The agent does not stop to ask before it reads a file, writes a file, deletes a file or runs a command. It reaches every file your login reaches, and nothing outside it. You accept that when you press Run.
This stands on the first page because it is a thing to decide before you install, not after. The licence terms say it in full.
The desk moves no money. It sends nothing outbound over the daily cap. It acts on nothing below the confidence floor. It never approves its own work.
These four hold on every run, whatever the agent asks for. How it works says what each one stops.
Two containers and one command: the desk and its database. A third takes a backup every night and encrypts it with a passphrase only you hold. You open the desk in a browser at an address on your own network.
Nothing leaves your machine. Your mail, your tasks, your notes and your database stay on the machine you installed this on. There is no account to make with us, and no server of ours that your desk depends on. The privacy statement says what we hold, which is your invoice and one row for your key.
Add the desk to your phone's home screen. It opens full screen with its own icon, and every page is two taps away. When the signal drops, the last screen you opened still reads.
The desk is published on your own machine and on no network. To reach it from outside the building you put a free tunnel in front of it — Tailscale or Cloudflare, about ten minutes, and the install page walks through both. We run no server for this, so there is nothing of ours between your phone and your desk.
Claude Code is Anthropic’s program. Codex is OpenAI’s. You install one under your own account, and you pay that company directly. What it costs is between you and them.
Your key gives you one year of updates and one year of answers to your questions. A person who reads, approves or edits is free on every tier, on every machine. There is no seat.
When your key lapses nothing switches off: you keep every release dated before the lapse, and only new releases and our answers stop. The price page carries the four tiers, the trial and the ceilings each one allows.
Ask within fourteen days of your first purchase and we return the money in full, even after you have downloaded and installed it.
The trial is everything in Team, for seven days. No card, and one trial per address. The install takes about ten minutes. You need Docker and a mailbox you can connect.