Linux Systemd Agent Deployment — Setup Guide

Prerequisites

• Linux system with systemd (Ubuntu 20.04+, Debian 11+, Fedora 35+)
• OpenClaw / Clawdbot / Hermes agent installed
• sudo access (for system-level services) or user systemd (systemctl --user)

Capabilities

Capability	Description
Unit File Generation	Create systemd .service files for agent processes
Restart Policies	on-failure, always, no — with RestartSec delay
Journald Logging	stdout/stderr automatically captured to journald
Resource Limits	MemoryMax, CPUQuota, TasksMax for agent processes
Environment Variables	Inject API keys, tokens, config paths via Environment=
User vs System Services	Choose user-level (systemctl --user) or system-level

Installation

npx skills add alphaonedev/openclaw-graph/linux-systemd

Quick Start

User-Level Service (Recommended for Agents)

Create ~/.config/systemd/user/hermes-gateway.service:

[Unit]
Description=Hermes Agent Gateway
After=network-online.target
Wants=network-online.target

[Service]
Type=simple
ExecStart=/usr/local/bin/hermes gateway run --profile corpusiq
Restart=on-failure
RestartSec=30
MemoryMax=4G
CPUQuota=200%
Environment=HERMES_HOME=/home/hermes/.hermes/profiles/corpusiq
Environment=PATH=/home/hermes/.hermes/node/bin:/usr/local/bin:/usr/bin:/bin

[Install]
WantedBy=default.target

Enable and start:

systemctl --user daemon-reload
systemctl --user enable hermes-gateway.service
systemctl --user start hermes-gateway.service
systemctl --user status hermes-gateway.service

System-Level Service

For agents that need to start at boot before user login:

[Unit]
Description=Hermes Agent Gateway (System)
After=network-online.target
Wants=network-online.target

[Service]
Type=simple
User=hermes
Group=hermes
ExecStart=/usr/local/bin/hermes gateway run --profile corpusiq
Restart=on-failure
RestartSec=30
MemoryMax=4G
CPUQuota=200%
Environment=HERMES_HOME=/home/hermes/.hermes/profiles/corpusiq

[Install]
WantedBy=multi-user.target

Install to /etc/systemd/system/hermes-gateway.service and:

sudo systemctl daemon-reload
sudo systemctl enable hermes-gateway.service
sudo systemctl start hermes-gateway.service

Key Configuration Patterns

Restart Policy Selection

Policy	Use When
no	One-shot tasks, cron jobs
on-failure	Long-running agents (prevents death loops from clean exits)
always	Critical services that must never stop (use with caution)

Resource Limits

MemoryMax=4G          # Hard memory cap — agent gets OOM-killed if exceeded
MemoryHigh=3G         # Soft limit — agent gets throttled
CPUQuota=200%         # Max 2 CPU cores
TasksMax=512          # Max threads/processes

Viewing Logs

# User service
journalctl --user -u hermes-gateway.service -f

# System service
journalctl -u hermes-gateway.service -f --since "10 minutes ago"

CorpusIQ Use Cases

1. Gateway Service: The CorpusIQ Hermes gateway runs as hermes-gateway-corpusiq.service on DGX Spark. Restart=on-failure prevents death loops from CLI --replace conflicts.

2. Cron Workers: Long-running monitor workers (email, social) can run as systemd services instead of cron, giving better logging and restart behavior.

3. Multi-Agent Deployments: Each agent profile gets its own systemd unit — isolation via MemoryMax and separate environment variables.

Troubleshooting

Symptom	Likely Cause	Fix
Service in activating (auto-restart) loop	Gateway crashing on start	Check journalctl --user -u <service> -n 50 for error
ExecStartPost exits code 1	Post-start script failing	Add || true guard or fix the script
OOM kill in logs	MemoryMax too low	Raise MemoryMax or investigate memory leak
Service stops after SSH disconnect	Using system service without linger	Run loginctl enable-linger <user>

---

← Skills Catalog | Marketplace → ↑ Skills Home

---

Curated by CorpusIQ — one MCP endpoint, all your business tools.