For teams.

A software factory your security team can read: agents run on machines you control, on the accounts you already have, with nothing of ours in between.

What a team gets today.

Magnum is free, and every developer runs it on their own Mac or Linux machine. These hold for each of them.

Your infrastructure

Agents run on your Macs, your servers and your cloud accounts: Docker Sandboxes on your Docker account, or any VM you can SSH into. Magnum’s node runs as the user, with no sudo and no open ports. Code and prompts go only to those machines and the model providers you choose.

Your accounts

Each agent uses its own sign-in: Claude plans or API keys, ChatGPT, GitHub Copilot, Google, or your own Anthropic-compatible endpoint. Several Claude accounts can be added and picked per project. Keys stay in Keychain or Secret Service.

Your rules

Allow, ask and deny rules per project, commands in Claude Code’s OS sandbox, the repository’s own check before any worker lands, and what each agent spends by day, model and project.

The short version.

What your security review will ask first. The privacy page has the rest.

AccountNone
Our serversNone between you and your agents. The iPhone’s relay passes only sealed messages.
Code and promptsYour machines, your model providers and, sealed, an iPhone you pair
TelemetryOne anonymous event as the app opens; off in Settings
On each machineOne binary in ~/.magnum, as the user; no sudo
UpdatesSigned; each machine’s node follows in place
PriceFree

Building for teams.

Magnum is used one developer at a time today. Shared machines and swarms across a team are being planned. Tell us what yours needs.