Short answers.

A free app with a team of AI bots. Each bot has a job, its own browser and a memory you can see. They hand work to each other, run routines and ask you before anything important.

Hesperan Bots is free. Your model provider’s terms, limits and costs apply; a local model costs nothing. With an API key, the app shows what each task cost.

It works and we use it every day, but it is still young. If something goes wrong, tell us from the app with Help → Report a Problem. New versions arrive automatically. Your bots, chats and memory stay as they are when the app updates.

A free Hesperan account gets you the downloads and updates, and lets your iPhone reach your bots away from home. Your bots, chats and memory never go to it.

A Mac with Apple silicon and macOS 13 Ventura or later, or a PC with Windows 10 or 11 (64-bit, x64). Bot computers need a Mac with macOS 26. Windows SmartScreen warns before the first start because the installer isn’t code-signed yet: choose More info, then Run anyway.

Yes, on your own server: a Linux computer that stays on, such as a home server, a NAS with Docker, a Raspberry Pi 5 or a rented server (x86_64 or ARM64). Your bots keep working while your Mac is closed, and you use them from the Mac app and the iPhone app. Server guide

Almost nothing. Your ChatGPT or Claude plan works there too: install Codex or Claude Code on the server and sign in there once (the one-click sign-in from the Mac app doesn’t reach a server yet). The bots’ own computers need a Mac, a server can’t use your Mac’s folders, and moving a Mac’s bots to a server comes later. What works on a server

Yes, in public testing on TestFlight. It pairs with your Mac or your server: at home directly, away from home through hesperan.com, encrypted end to end. It tells you when a bot needs you.

Any model you like, and each bot can use a different one: your ChatGPT or Claude plan (through the Codex app or Claude Code on your Mac, signed in with your own account), an API key for OpenAI, Anthropic, OpenRouter or any OpenAI- or Anthropic-compatible service, or a local model in Ollama or LM Studio. On a server, install Codex or Claude Code there and sign in once to use your plan. You don’t need a new subscription.

Before a bot buys, pays, orders, subscribes or sends something to other people, in its browser or on its computer, it stops at the last step and asks. Spending money is off for every bot until you allow it. When you do, each purchase or payment still waits for your OK unless you choose Always for that bot; with Never, no rule can turn it back on. Sending asks first by default and can be set to Always for a bot, and mail through a connected account arrives as a draft. The app recognises buy and send buttons from the page, which works on common sites but can miss unusual ones, so only sign a bot in where you would trust it to act.

The internet, if you allow it, and one shared folder, if you want one. Not your Mac, your home network or your other devices. You can watch it and take over at any time. It is optional, needs macOS 26 on a Mac with Apple silicon, and takes about 1.5 GB of memory while it runs.

One button, or ⌘ and full stop on the Mac, stops every bot at once. File changes can be undone for 30 days.

Chats, memory, files and settings are stored on your Mac, or on your server if your bots run there. To answer, a bot sends the conversation and what it needs from memory to the model provider you chose; with a local model, that stays with you too. Websites and services you let a bot use see what it sends them. There is no analytics or telemetry in the app: a problem report reaches us only when you choose to send one. Privacy policy

No. Away from home your devices connect through hesperan.com, encrypted end to end: it passes the data on but can’t read it. hesperan.com only handles your account, downloads, updates and this relay.

The app is licensed under AGPL-3.0-only. Its source is private for now. Programs by others inside it keep their own licences. Source code and licences