p

A thin, friendly wrapper around systemd. Run long-lived processes with a CLI that feels like pm2, but every service is a real systemd unit you can poke at with systemctl whenever you want.

Run p <command> --help for the full flag list of any subcommand.

Install

go install github.com/ccev/p@latest

Requires Go 1.25+, systemd, and journalctl. Run as your user (units land in ~/.config/systemd/user/) or as root (system-wide units).

Update

Same command — go install always fetches the latest tagged version and overwrites the binary in $GOBIN (usually ~/go/bin/p):

go install github.com/ccev/p@latest

Existing services keep running on the old binary's unit format. To re-render one through the current template (e.g. to pick up the login-shell wrapping), do any declarative edit:

p edit <name> --cmd "<same or new command>"

Examples

Start something

$ p start "node app.js" -n web
● web
  unit   /home/me/.config/systemd/user/p-web.service
  cmd    node app.js
  cwd    /srv/web
  state  started

See what's running

$ p status
id  name    status      pid     uptime   ↺   cpu    mem      net ⇣/⇡
0   web   ● online      12345   2d 3h    0   1.2%   142.0MB  4.1MB/1.2MB
1   worker● online      12380   2d 3h    2   0.4%   58.0MB   —
2   bot   ● failed      —       —        7   —      —        —

3 service(s)  2 online  0 stopped  1 failed

p status -w keeps it refreshing. Columns drop out on narrow terminals so it still reads on a phone.

Tail the logs

$ p logs web
2026-06-15T09:42:01+0000 p-web.service[12345]: listening on :3000
2026-06-15T09:42:09+0000 p-web.service[12345]: GET / 200 12ms
2026-06-15T09:42:11+0000 p-web.service[12345]: warning slow query (134ms)

-l 200 for more scrollback, --grep to filter, pass multiple names to multiplex with per-service color prefixes.

Restart, stop, reload, delete

$ p restart web
● web restarted

$ p reload web                # SIGHUP — re-read config without restarting
● web reloaded (SIGHUP → pid 12345)

$ p stop web
● web stopped

$ p delete web
✖ web deleted

Logs are size-capped per service

Each service's journal is isolated and capped at 20MB by default — older entries get rotated out automatically, no cron needed. Override per-service with --log-max (e.g. --log-max 100M, --log-max 1G), and clear a service's logs on demand with p flush:

$ p flush web
● web logs flushed

$ p flush --all

Inspect or edit a service

$ p show web
● web  online
  description  p-managed service: web
  unit         /home/me/.config/systemd/user/p-web.service
  pid          12345
  uptime       2h13m
  memory       142.0MB
  cpu          1.2%
  command      /bin/sh -c 'node app.js'
  cwd          /srv/web
  restart      always (after 5s)

$ p edit web --memory-max 512M --env DEBUG=1
● web edited & restarted

p edit web with no flags opens the unit in $EDITOR.

Suggested zsh aliases

alias pst='p status'
alias pl='p logs'
alias pr='p restart'

---

This repository was written entirely by an AI agent.