Compare
Belvador vs self-hosted Sentry
Both keep error data on infrastructure you control. Self-hosted Sentry gives you the full Sentry stack, and you run all of it. Belvador gives you a smaller product on one small server that Belvador creates, updates and backs up inside your Hetzner project.
The short answer#
Both keep error data on infrastructure you control.
Self-hosted Sentry needs 4 CPU cores and 16 GB of RAM plus swap, and you run and upgrade its Docker Compose stack.
Belvador runs one small node in your Hetzner project and handles updates, backups and restores for you.
Self-hosted Sentry has the full Sentry feature set except Seer and other AI features; Belvador has fewer features.
Self-hosted Sentry runs anywhere; Belvador's node runs on Hetzner Cloud only.
Side by side#
| Belvador | Self-hosted Sentry | |
|---|---|---|
| Minimum server | One CX23: 2 shared vCPU, 4 GB RAM, 40 GB disk | 4 CPU cores, 16 GB RAM + 16 GB swap (32 GB recommended), 20 GB free disk |
| What you run | No Docker Compose to run; Belvador applies node updates and backups | A Docker Compose stack; the compose file defines 53 services |
| Upgrades | Rolling node releases applied by Belvador; you can pause them | Monthly releases; old versions are not patched; upgrades stop services and run migrations |
| License | Hosted service; SDK packages are MIT | Functional Source License; Apache-2.0 after two years |
| AI features | No AI root-cause analysis; an MCP server (remote with OAuth, or local) lets your own agent read issues, traces, logs and monitors | Seer and other AI/ML features are not included |
| Where data lives | On a server in your own Hetzner Cloud project; Belvador provisions and updates it | Wherever you run it |
Self-hosted Sentry column: from their public pages, read on Oct 25, 2026. Prices and limits change — check the source:
Belvador column, from the docs: Connect Hetzner, Update and delete, Choose an SDK, MCP for AI agents, Data protection.
What running self-hosted Sentry involves#
Sentry's self-hosted guide asks for 4 CPU cores, 16 GB of RAM plus 16 GB of swap, and 20 GB of free disk; 32 GB of RAM is recommended. The stack is a Docker Compose file with dozens of services. Sentry releases monthly, does not patch older versions, and an upgrade stops the services and runs migrations. None of this is hard for a platform team, but it is a standing job.
Self-hosted Sentry also leaves out Seer and other AI features, and it is licensed under the Functional Source License, which turns into Apache-2.0 two years after each release.
What Belvador does instead#
One node. A CX23 (2 vCPU, 4 GB) in Falkenstein, Nuremberg or Helsinki, created in your Hetzner project with a firewall and a hardened host (connect Hetzner).
Signed updates. The node's agent checks for signed releases every 15 minutes, takes a backup first and rolls back a release that fails to start (updates).
Daily encrypted backups to Cloudflare R2 with a 7-day deletion lock and a weekly restore drill (backups).
No SSH needed. Resize, replace-and-restore and delete are buttons on the Node page.
What Belvador does not do#
The full Sentry feature set: no replay, profiling, Seer or the integration catalogue.
Root access as the normal way of working. You own the Hetzner project and can see the server, but the node software is Belvador's and is managed through the app.
Other clouds. Today the node runs on Hetzner Cloud only (BYOC on Hetzner).
Questions#
Can I move from self-hosted Sentry without changing code?#
Yes: point the DSN at Belvador. Old events stay in your Sentry install; Belvador does not import them (migrate from Sentry).
Who can read the data on the node?#
People in your workspace, through the app. Belvador's control plane stores settings and counts, not event bodies (data protection).
Posts on this topic
All postsTry it on your own Hetzner project
Free during the beta. Hetzner bills the server; Belvador never bills per event.