Run your own Matrix community
The software’s free. The 3am wake-up when an update goes sideways is the part nobody prints on the box. Two ways to run a Matrix community: yourself, or through us. This is what each one really costs.
MatrixRooms.info runs on the same open playbook anyone can use to self-host Matrix: matrix-docker-ansible-deploy. etke.cc is the managed version of it. Same stack, same code. The only question is who runs it: you, or us.
Who owns what
Getting it running
You self-host
A server, a domain, DNS you control, and root over SSH. Then enough Ansible to drive the playbook.
etke.cc runs it
You order, answer a few questions, and it shows up built: Docker, the homeserver, backups, firewall, fail2ban, SSH hardened.
Weekly updates and security
You self-host
You re-run the playbook yourself. It does not update while you sleep.
etke.cc runs it
We ship server and Matrix updates every week, whether or not you remembered they were due.
Breaking changes
You self-host
The playbook stops and makes you read every breaking change before it will run again. A real safeguard, and also the reason you end up reading release notes at midnight.
etke.cc runs it
We read the changelog so you don't. It rides along in the weekly pass.
Backups
You self-host
Yours to set up: a database dump and a copy of the files, on a schedule you build and hope you got right.
etke.cc runs it
Nightly, kept a week. And if a disk dies, we're the ones doing the restore, not you.
Monitoring and uptime
You self-host
Prometheus and Grafana are in the box. You still hook them up and stare at the graphs while your coffee goes cold.
etke.cc runs it
Monitoring that pings us the moment your server flatlines, email and Matrix both.
When it breaks at 2am
You self-host
You're on call. It's your server.
etke.cc runs it
You send us a message. We're already up. Pulling servers back from the brink is sort of our whole thing.
Your time
You self-host
As much as it takes. Some months an evening, some weeks a bad one.
etke.cc runs it
Ours, not yours. That's the thing you're actually paying for.
What self-hosting actually takes
If you already run Linux boxes, the playbook is one command. What comes after, the updates, the changelogs, the 2am wake-ups, that’s the part you actually host. The playbook’s own README doesn’t oversell it: it “tries to make self-hosting and maintaining a Matrix server fairly easy. Still, running any service smoothly requires knowledge, time and effort.” That last sentence is the whole ballgame.
Here’s what stays on your plate:
- A server, a domain, DNS you control, and root over SSH.
- Comfort on the command line, and enough Ansible to run the playbook.
- Re-running upgrades yourself. Nothing updates on its own.
- Reading breaking changes. The playbook halts and won’t run again until you acknowledge each one, so you don’t sleepwalk into an upgrade that eats your config. The changelog you’re acknowledging is past 4,600 lines and grows every week.
- Your own backups, and Postgres major-version upgrades by hand.
You’ve likely done all of it before. It’s just real, recurring, and yours. The playbook is on GitHub, and it’s the real thing.
It’s a stack, not a box
“A Matrix server” is shorthand. Every server here is a pile of services wired to work together, and each one ships already configured:
- Synapse, the Matrix homeserver itself
- a web client, your pick of Element, Cinny, or FluffyChat
- Ketesa, a web admin panel for the whole thing
- coturn, so voice and video calls actually connect
- PostgreSQL, tuned and backed up nightly
- Traefik, with Let's Encrypt certificates that renew themselves
- monitoring that pages both you and us when something slips
That’s just the base, and if you run it yourself, every line of it is yours to wire and keep patched.
Add-ons, if you want them
On top of the base there are more than forty optional components. Turn on what you want, skip the rest, pay for what you turn on.
- Bridges to fifteen networks people already live on: WhatsApp, Signal, Telegram, Discord, Slack, Instagram, Messenger, Google Chat, IRC, and more. The folks on the other side never have to touch Matrix.
- Six bots: a private conversational AI (baibot, Venice by default, so your chats don’t feed someone’s training set), Draupnir to keep the trolls out, Honoroit for a helpdesk, plus reminder, web-form, and plugin bots.
- Around twenty extra services for what a community actually reaches for: Jitsi for video, Etherpad for shared docs, email hosting, single sign-on, a password manager, and more.
Two of them are ours, and they’re why a managed server doesn’t feel like a raw one:
- Ketesa, the admin panel: bulk user import with a dry-run first, per-room storage stats, one click to reclaim disk from old media, a live status page, billing and support in the same window. Vanilla Synapse gives you none of that.
- The Scheduler: run or schedule maintenance by messaging a bot, get pinged when a disk fills or an endpoint goes dark, and see exactly what each component is costing you.
Self-hosting, every one of these is another container to stand up and keep current. Managed, it’s a checkbox that rides the weekly update pass with everything else.
What etke.cc actually does
Pick a size, from Small (2 vCPU, 2 GB) to Monster (16 vCPU, 32 GB), on a box we rent for you or your own, in one of six regions. Outgrow it and you bump up later.
Then we keep it boring, which is the entire job. Updates land weekly, backups sit for seven days with a self-service restore, and monitoring pages both of us the moment a port, a DNS record, or a Matrix endpoint goes dark. Setup ships hardened: fail2ban, a firewall, locked-down SSH, a tuned database.
A couple of honest caveats, because you should hear them from us. The weekly pass doesn’t cover big OS jumps, the kind where the whole distribution moves up a release; those we do on request, not on autopilot. And if you hand-edit the server config, the automation quietly puts it back, because predictable is what keeps it up.
None of this is exciting. It’s a server that stays up and a team that answers when it doesn’t.
No lock-in, and we mean it specifically
- Every line is open source (AGPL): the upstream playbook and etke.cc’s own components alike. You can read exactly what runs on your server before it runs.
- On a Bring-Your-Own-Server plan you hold root the whole time. It was always your box, so there’s nothing for us to hand back.
- The subscription is monthly. Cancel before the next renewal and that’s it: no contract, no exit interview.
In etke.cc’s own words: “If you ever become unhappy with our service, you can always host your server elsewhere or start maintaining it yourself.” The playbook to do exactly that is the same one linked above.
Done reading changelogs before you’ve even started?
Rather run it yourself? The playbook’s on GitHub. No hard feelings. We build it in the open for exactly that reason.