Support and maintenance
You can fire me and keep the stack.
What support means when the infrastructure is yours — advisory, not remote hands.
What you own
The stack runs on your server, under your account, on hardware you rent from whoever you like. The components are open source; the generator and the rights service are source-available under a licence that lets you read, build, modify and run them internally for nothing. The generated configuration is plain Docker Compose that you can read, diff, and change.
I do not have access to it. I do not want access to it. That is not a limitation of the support offering — it is the product working correctly.
About the warranty
There isn’t one. Every component in the stack — Authentik, Gitea, Caddy, Woodpecker, Vaultwarden, Postgres — already ships under a licence disclaiming warranty, and that is true whether you get them from me or install them yourself. Curating and integrating them does not change the underlying position, and a vendor telling you otherwise would be selling you something that does not exist.
Worth checking what you have today before you treat that as a downgrade. A managed provider’s SLA is not a warranty of correctness. It is a promise to credit back a percentage of one month’s fee if availability drops below a threshold. It does not restore your data, cover your downtime, or bring your customers back. Most teams have never read theirs.
So: not less protection than you have now. A clearer arrangement, where you hold the infrastructure and can buy my attention by the month.
You can fire me and keep the stack.
You cannot fire your cloud provider and keep the platform.
What “help when you need it” actually means
I am an advisor, not an operator. You keep the keyboard. I never log in to your server.
That sounds like a limitation until you look at what actually goes wrong. Almost every real problem is diagnosable from artifacts you can send me in a message: the failing log, the Compose diff, the error output, the migration that stopped halfway. I have seen the stack fail in more ways than you have. Reading those artifacts is the work.
Concretely, this is what you get:
Diagnosis from artifacts. Send logs, config diffs, and error output. A collector script in the stack assembles the configuration part for you as a secrets-free snapshot — its redaction fails closed, so anything it cannot prove is safe never leaves your server. You get back an explanation and the specific change to make. You apply it, on your schedule, with your hands.
Divergence review — the one that matters most. You will customise the stack. That is expected and it is why it is generated rather than hosted. But every change you make is a change an upstream release can collide with. Send me the collector’s snapshot before an upgrade — I regenerate your configuration from the same CLI release and diff the two, so drift shows up even where nobody remembers making a change — and you get back what will break and what to do about it. This is proactive, it takes me an hour, and it converts the worst failure mode — a broken upgrade at 22:00 on a Tuesday — into a planned piece of work. Maintained’s included advisory hour covers one review when you choose to spend it there; Priority makes it standing practice before every upgrade.
Pairing when it’s serious. Screen share, you drive, I advise in real time. Your credentials never leave your machine, no access is granted, and the fix is applied by the person who will have to maintain it.
What I will not do: log in, hold your credentials, apply changes on your behalf, or take responsibility for your production environment. If a vendor offers that at this price, ask what happens to your DPA when they do — they have just become a processor with access to your systems, which is a rather large thing to acquire in exchange for a support ticket.
Tiers
Self-supported
Free
public repository — publishing now
- Full source and documentation
- Issue tracker for defects in the software
- Community discussions
- No response commitment
Maintained
€249
per month, per production stack
- Tested upgrade paths for shipped components
- Security advisories filtered for your configuration
- Async questions, response within three working days
- One advisory hour included monthly
Priority
€749
per month, per production stack
- Everything in Maintained
- Next working day response
- Three advisory hours included monthly
- Pre-upgrade divergence review
- Scheduled pairing sessions
Additional advisory time is €135 per hour. All prices exclude VAT. Monthly billing, cancel whenever it stops earning its place. There is no annual lock-in, because asking a bootstrapped team to commit for a year to a one-person vendor would be a bit rich.
Scope
support-scope
stack.generated
The stack as generated, and the integration between its components.
stack.upgrades
Upgrade paths for the components shipped in the stack.
stack.diagnosis
Failures in the stack layer, diagnosed from artifacts you send.
config.divergence
Review of your changes against the generated baseline, and what upstream releases will do to them.
your.application
Your own code. Different product, different conversation.
your.server
Hardware, hosting provider, network, operating system.
your.data
Your data itself — what to restore, when, and living with the result. The stack drills a restore monthly; reading and signing off those reports is yours.
components.added
Anything you bolt on that the stack does not ship.
access.privileged
Credentials, logins, and changes applied on your behalf. Not offered at any tier.
The exclusions are not there to protect me from you. They are there so that neither of us discovers a disagreement about scope during an incident.
About the person
One person, operating through a Dutch BV in The Hague, Netherlands. The career detail is on the about page; what matters here is the shape of the promise.
No support desk, no rota, and no overnight cover. I work 09:00–17:00 CET, Monday to Friday, I take holidays, and when I am away it is posted on this page in advance. If you need a follow-the-sun operation, buy one — you will not be happy here.
What you get instead is that the person answering has read every line of the integration and will still be the same person in eighteen months. If you have been through an enterprise support queue, you already know why a named human with a smaller promise is worth more than a large one nobody keeps.
The free tier, and what it isn’t
The source is open to read because a stack you cannot inspect is not a stack you own. The rights service is being published as a public repository alongside the write-up series in Insights.
The issue tracker is for defects in the software. It is not a deployment helpdesk. If something in the stack is broken, file it — that is useful to me and I want it. If you need help getting your particular server, network, and customisations working together, that is the paid product, and I will say so politely and close the thread.
No hard feelings either way. Running it yourself for free is a completely legitimate thing to do — the licence explicitly allows it, and the readable source is what makes it possible.