Two ready-made n8n workflows for people who run their own servers and their own n8n: one patches a Proxmox container and rolls it back if it breaks, the other explains your failed n8n runs in plain English. They run in your own n8n and you keep them. One-time purchase, no subscription.
Proxmox Safe Patch $59
Monthly updates for a Proxmox container without babysitting it. Safe Patch checks your app answers, takes a snapshot, runs apt upgrade inside the guest, then checks the app again. Still working: the snapshot stays as a rollback point and older ones are cleaned up. Broken: the guest is rolled back to the snapshot, started again, and you get an alert saying what failed.
Report-only (dry run) until you switch it on
Never patches a guest that is already failing a check, and never installs without a finished snapshot
Only deletes snapshots it made itself, and only after the check passed
Reboots only if you allow it, followed by a second check
Optional history table in your own Postgres (off by default)
Tested on Debian 12 and Ubuntu 24.04 containers, including a reboot. Virtual machines use the same steps but are not tested yet. Good for web and app servers, reverse proxies and static sites. A rollback loses anything written inside the guest since the snapshot, so use it on databases only if they have their own backups.
Needs: n8n that can reach your Proxmox API on port 8006 (Proxmox VE 9 (tested on VE 9)), a Debian or Ubuntu guest that n8n can reach over SSH (the setup guide creates a user that may only run apt-get), and snapshot-capable storage. It talks only to your own Proxmox API, to the guest you choose, and to the health-check URLs you list.
40 seconds: a dry run, a forced rollback and a real patch on a throwaway test container. Real run data; voice-over is AI; network names are swapped for neutral labels.
Watches your own n8n for failed runs, checks simple rules first, and only asks Claude (Anthropic's AI) when the rules are not enough. You get one clear alert per failure: what broke, the likely cause, a suggested fix, and whether a rerun looks safe.
Credential problems are flagged straight away with the step that failed
Retries only the workflows you list, a limited number of times. Nothing is retried by default
A monthly spend cap on the AI step, a limit per run, and at most one AI call per failure
Passwords, keys, tokens and email addresses are scrubbed before anything leaves your n8n
The AI only advises. Nothing it says is ever run
Needs: n8n with the public API turned on (all self-hosted versions; on n8n Cloud, Starter, Pro and Enterprise, not the free trial) and your own Anthropic API key, which you pay Anthropic for directly. The AI step sends Anthropic a redacted summary of each failure; the exact list is in the privacy note that comes with the download.
35 seconds: a failed run explained, a temporary error retried and healed, and a failure that kept failing. Real run data from a self-hosted n8n; voice-over is AI; names are swapped for neutral labels.
Check a download before you import it: sha1sum file.zip (or sha256sum; on Windows, Get-FileHash -Algorithm SHA1 file.zip) should print the value below. The same list as a text file: CHECKSUMS.txt.
Pay securely. Your card details go to the payment processor, never to us.
Your download link arrives by email within a few minutes.
Import the workflow into n8n, create the credentials the setup guide lists, fill in the Set Configuration Parameters node, and run it once.
You set it up and run it yourself using the setup guide in each download; see the terms of sale. Looking for the Proxmox backup, audit and report tools? They are on the Proxmox tools page. Questions before you buy: [email protected].