Machines

A machine in Mantler is any hardware — local workstation, cloud VM, bare-metal server — that has mantlerd installed and running.

What mantlerd does

mantlerd is the open-source daemon that runs on each worker machine. It:

• Checks in with the Mantler control plane periodically via HTTPS
• Executes typed commands sent by the control plane (pull a model, start a runtime, etc.)
• Reports hardware capabilities (GPUs, RAM, installed runtimes, deployed mantles) back to the control plane
• Relays inference traffic between external clients and local runtime backends

The daemon communicates only with the server URL you configure. By default this is the Mantler hosted service, but you can point it at any compatible control-plane URL, including a self-hosted one.

The full source is available at github.com/Borgels/mantlerd.

Architecture boundary

mantlerd is the public, open-source half of the Mantler system. The control-plane server is separate and private. This separation is intentional — you can read every command type, execution path, and network payload in the daemon's repository.

See the Security & trust model page for a full account of what the control plane is and is not allowed to ask the daemon to do.

In this section

• Install — install mantlerd on Linux or macOS
• Configuration — config file, CLI config commands, trust mode
• CLI reference — all mantler subcommands
• Security & trust model — command allowlists, audit log, release verification