Runtime · Open Source

A runtime for
agents that stay.

NockOS is the open-source local runtime for long-lived AI agents: process supervision, heartbeats, and uptime tracking. One of many runtimes Nock Command coordinates, not the only one.

Install NockOS View on GitHub ↗

$ brew install nocktechnologies/tap/nockos

What it runs

Keep agents
alive and honest.

NockOS is a single binary that supervises agent processes on your own machine. No daemon spread, no cloud.

Supervision 01 · proc

Agents that crash are restarted in under a second. Crash loops are caught and reported, not hidden.

[supervise]
restart  = "always"
backoff  = "1s"

Heartbeats 02 · beat

Every agent emits a heartbeat. Miss two and NockOS flags it: you know before the work stalls.

[heartbeat]
interval = "10s"
miss_max = 2

Uptime 03 · log

Per-agent and fleet-wide uptime, logged locally. Pipe it into Nock Command for one-screen visibility.

[uptime]
track  = "per-agent"
export = "command"

How it works

Run a fleet.
In four steps.

NockOS is a single binary. Point it at your agents and it keeps them running.

Install the runtime

A single binary via Homebrew or cargo. No daemon, no kernel module.

Register your agents

nockos register names each agent and its start command. One line per agent.

NockOS supervises them

Crashes restart, heartbeats are checked, uptime is logged automatically.

Pipe status into Command

nockos export streams fleet status to Nock Command for one-screen visibility.

Install

One line. No accounts. No telemetry.

# homebrew (macOS, Linux)
$ brew install nocktechnologies/tap/nockos

# initialize the runtime
$ nockos init

# start supervising
$ nockos start
[ok] runtime online · 5 agents supervised