Lost Your Hermes Agent Data After Months of Use? This Backup Tutorial Can Save You

Used Hermes Agent for a few months, and your data's gone? This backup guide can save you. Let's start with three real-world scenarios—see if any of these have happened to you:

Scenario 1: Reinstalling the OS

Your computer gets slower and slower over time, so you decide to reinstall the OS for a fresh start. After setting up the system and installing your software, it suddenly hits you—what about all those months of stuff you had in Hermes?

The naming habits you corrected dozens of times, the things you told it

💡 What You Will Learn

Used Hermes Agent for a few months, and your data's gone? This backup guide can save you. Let's start with three real-world scenarios—see if any of these have happened to you: Scenario 1: Reinstallin

Used Hermes Agent for months and lost all your data? This backup guide can save you

Three real-world scenarios — see if any of these have happened to you:

Scenario 1: Reinstalling your OS

Your computer has gotten slower and slower over time, so you decide to wipe it clean with a fresh OS install. After reinstalling the system and your software, it suddenly hits you — what about all those months of stuff you had in Hermes?

The naming conventions you corrected dozens of times, the project structure you taught it, the dozen-plus skills it auto-generated... all of it was on the old system drive. And you've already formatted it.

Scenario 2: Your hard drive suddenly dies

SSDs are more reliable than mechanical drives, but they're not invincible. One day you boot up, the system won't load, the drive won't even register. The Agent memory bank you spent months training goes down with that drive — physically gone.

Call a data recovery service? They'll quote you thousands just to get started, and there's no guarantee they can even recover it.

Scenario 3: Switching to a new computer

You finally upgrade to a new machine — faster, bigger screen. But you don't want to start from scratch with Hermes. You just taught it your coding style a few weeks ago. Do you really want to do that all over again?

Sure, you could pull the drive out of your old computer — but what if you already sold it? What if it's a company laptop you have to hand back when you leave?

All three scenarios boil down to one question: does your Hermes data exist independently of the machine you're currently using?

If the answer is no, you're running naked.

Hermes has three core assets you need to back up: memory bank (personal preferences and context, accumulated over months), skill library (auto-generated skill files, each one fine-tuned through multiple rounds of conversation), and config files (model, tool, and permission settings). On Windows they live in C:\Users\YourUsername\AppData\Local\hermes\, on Mac/Linux in ~/.hermes/.

Below are two approaches — pick whichever fits your situation, or go with both for double insurance.

Cloud backup: no fear of dead drives, no fear of lost computers

Storing your data in a private GitHub repo is like having a remote vault. Even if your computer gets stolen or your drive burns out, your data survives.

Manual version: set it up once, benefit forever

cd ~/.hermes
git init

Create a .gitignore to keep sensitive files out:

logs/
cache/
.env
.env.*
# sessions/    # Optional — session logs, skip if you have too many files

Add the core data:

git add config.yaml memory/ skills/ auth.json profiles/
git commit -m "First backup: complete Hermes Agent data"

Create a private repo on GitHub (must be Private, not Public), then push:

git branch -M main
git remote add origin https://github.com/YourUsername/hermes-backup.git
git push -u origin main

From now on, backing up is just three commands:

cd ~/.hermes
git add -A && git commit -m "Backup $(date +%Y-%m-%d)"
git push

Automated version: set it once, never think about it again

Write an auto-backup script backup-hermes.sh:

#!/bin/bash
cd "$HOME/.hermes"
if git status --porcelain | grep -q .; then
    git add -A
    git commit -m "Auto backup $(date '+%Y-%m-%d %H:%M')"
    git push origin main
fi

Then set up a scheduled task:

• Mac/Linux: add 0 9,18 * * * /path/to/backup-hermes.sh to crontab -e — runs twice daily, morning and evening
• Windows: use Task Scheduler, create a daily task that runs Git Bash with the script path
• Hermes built-in cron (if you use Hermes): one command does it

hermes cron create "0 9,18 * * *" \
  --prompt "cd ~/.hermes && git add -A && git commit -m 'auto backup' && git push origin main"

I recommend backing up at least before you start work and after you finish every day. If you make changes frequently, even hourly works.

Local backup: fast, no network dependency

Cloud backup is reliable remotely, but it depends on GitHub being reachable. Local backup is the most direct safety net — just copy the entire Hermes directory to an external drive or NAS.

Manual version: drag and drop, done

The simplest method: every few days, copy the entire ~/.hermes directory to an external drive or another computer.

• Windows: copy the whole C:\Users\YourUsername\AppData\Local\hermes\ folder to a USB stick or external drive
• Mac/Linux: cp -r ~/.hermes /Volumes/BackupDrive/hermes-backup/

To restore, just do the reverse — copy the backup directory back over the original.

Automated version: incremental backups with rsync or robocopy

The problem with manual copying is — you'll forget. Set up an auto script that only copies new and changed files:

Mac/Linux with rsync:

#!/bin/bash
rsync -av --delete \
  --exclude '.env' \
  --exclude 'logs/' \
  --exclude 'cache/' \
  "$HOME/.hermes/" \
  "/Volumes/BackupDrive/hermes-backup/"

Add to crontab: 0 12 * * 1 /path/to/backup-hermes-local.sh (backup every Monday at noon)

Windows with robocopy:

@echo off
robocopy "C:\Users\YourUsername\AppData\Local\hermes" "D:\hermes-backup" /MIR /XD logs cache .git /XF .env .env.*

Then use Windows Task Scheduler to create a daily task that runs this batch script.

The biggest advantage of local backup: no fear even if GitHub goes down — no rate limits, no queues, and you can restore from the external drive anytime.

My recommendation: do both

Cloud solves the "computer died" problem, local solves the "GitHub is down or no internet" problem. Run both simultaneously, and your complete backup strategy looks like this:

And these two approaches don't conflict at all — the GitHub script and the local rsync script can coexist, one running automatically, one on a schedule, covering each other's backs. Total config backup size is usually under 50MB, so it barely takes up any space whether cloud or local.

A lot of people treat AI Agents like disposable toys — don't want to manage them after use, don't care if the data is lost. But if you've spent weeks or even months fine-tuning it, not backing up is disrespecting your own time.

GitHub private repos are free, writing a backup script takes 5 minutes, and setting up a scheduled task is a one-time deal. With all that in place, the only reason not to back up is "nothing's gone wrong yet."

But by the time something goes wrong, it's too late.

Have you backed up your AI Agent yet? Cloud, local, or both? Share your strategy in the comments — and if you've got a better approach, feel free to share it so everyone can learn.