hermes recipes¶

working configs and scripts from my Hermes Agent walkthroughs. everything here ran on real hardware before it landed: a Mac Mini M4 that runs Hermes Agent 24/7, and an RTX 5090 Linux box for local inference.

Hermes Agent is Nous Research's open-source agent: autonomous skill creation, persistent memory with cross-session recall, cron jobs, MCP support, 20+ messaging gateways. docs.

the almanac¶

these recipes usually start as posts in the witcheer almanac, my weekly notes from running local AI on my own hardware. read the back issues or subscribe at buttondown.com/witcheer.

the bar¶

every command in a recipe ran on real hardware before publishing. no copy-paste from docs, no untested steps, no "should work". if a step has not been tested, it does not ship here.

recipes¶

24/7 Hermes Agent on a Mac Mini M4: the launchd gotcha and watchdog, the [SILENT] cron pattern, a git-synced workspace, the human-in-the-loop approval rule
Hermes Agent on a cheap VPS: the secure-box hardening script, the install one-liner, driving it over tmux, the systemd setup that survives a reboot
lock down a self-hosted Hermes Agent: what the gateway, dashboard, and API server expose, secret hygiene, backup/restore, least-privilege scoping, an exposure-audit script

Hermes Wingtips¶

the daily tip series, collected: Hermes Wingtips (65 tips, newest first). point an agent at llms.txt.

also here¶

threads & cards: Flightplan, heads-up cards, community pulse, builder spotlights
changelog: what shipped in Hermes Agent and how to use it, in my words

found a problem with a recipe? open an issue. I fix what I ship.