What is Belvador

Belvador catches the errors your apps throw — a crash in the browser, a failed server request, an exception you report yourself — groups them into issues and tells your team. Next to errors it keeps OpenTelemetry traces and logs, watches your cron jobs, checks your URLs and tracks SLOs. The difference from other error trackers: this data is stored on a server in your own Hetzner Cloud account, and Belvador sets up and maintains that server for you.

How it fits together#

your app ──(SDK + DSN)──▶ ingest.belvador.dev ──▶ your node, in your Hetzner project                                                        ▲your browser ──▶ Belvador app ──▶ edge.belvador.dev ────┘  signed read tickets

1. Your app sends errors with a Belvador SDK. The project's DSN tells the SDK where to send them.

2. The ingest edge at https://ingest.belvador.dev checks the DSN key and your limits, then forwards each error to your node. If the node is briefly unreachable, the edge keeps the data encrypted to your node's own key — Belvador cannot read it — and delivers it when the node is back.

3. Your node is a small server in your Hetzner project. It scrubs sensitive fields a second time, groups errors into issues and stores them.

4. The Belvador app shows issues, the Fleet overview and node health. Your browser reads error data straight from your node through edge.belvador.dev with signed tickets that expire within minutes. Belvador's own database only keeps account, project and configuration metadata.

What you do and what Belvador does#

You do not need SSH, Docker or a server login to get started. Setup and maintenance happen from the Belvador app.

What Belvador captures — and what it does not#

• Errors from the Belvador SDKs, traces and logs from OpenTelemetry. The 0.x SDKs send error events and cron check-ins; their own tracing, session replay, logs and profiling are dropped inside your app before anything leaves it. Traces and logs come from the standard OpenTelemetry SDKs with a project ingest key and are stored on your node as well. See Traces and logs.

• Automatic capture is what you wire up. In the browser, uncaught errors and unhandled promise rejections are captured as soon as the SDK starts. React error boundaries and root hooks, the TanStack Start router and middleware, and the Cloudflare Workers wrapper each capture their own part once added. See Choose an SDK.

• Handled errors stay handled. An error your code catches and deals with is not sent unless you call captureException yourself.

• Sensitive data is filtered twice — by the SDK and again on the node. Filtering works by field names and patterns, so it reduces risk but cannot recognise every secret. See Data protection.

What is available today#

Belvador is in beta and free during the beta; you pay Hetzner for the server. Belvador does not bill per event.

Where to go next#

• New here? Follow the Quickstart — from sign-up to your first issue.

• Installing in an app? Start with Choose an SDK.

• Setting up the server? Read Connect Hetzner and set up your node.

• Using an AI coding agent? Give it the Guide for AI agents.

• Something not showing up? See Troubleshooting.