Quickstart
Get an account, mint an API key, drop the agent into your docker-compose.yml, see logs and metrics flowing within ten minutes.
Lumin docs
One container, your whole stack. Logs, metrics, and the user sessions that cause them — in a single dashboard.
Lumin is a small monitoring stack for indie SaaS teams of one to five. The whole instrumentation story is one container next to your services plus one snippet in your browser app — no OpenTelemetry collector, no eleven product tabs, no per-GB roulette.
Start here
Section titled “Start here”Install the agent
The full agent reference: compose service, environment variables, Docker socket proxy, networking model.
Stacks
Per-stack snippets for Gin, Spring Boot, Express, FastAPI, and the browser SDK. Use the canonical http.* metric names so different services chart side by side.
Concepts
The mental model: how projects + API keys scope ingest, how logs are stored and live-tailed, how metric labels and cardinality work, how uptime checks probe from outside.
The five things Lumin tells you
Section titled “The five things Lumin tells you”- Is my site up? — HTTP checks probed from Lumin’s servers every 30 s to 15 min, with a per-check degraded threshold. The dashboard opens on them.
- Is my service healthy? — Host and container metrics from the agent, plus request rate, error rate, and latency from your own
/metricsendpoint. - What’s in the error log? — Structured logs with
{timestamp, level, service, message, attributes, trace_id}, live-tailable via SSE. - What are users doing? — Per-session timelines from
@lumin-monitor/browser(page/track/identify). - Who tells me when it breaks? — Threshold rules over one metric or a log pattern, delivered to Slack when they fire and again when they recover.
Tag your server logs with the browser’s session id and you can filter them by it — see Linking browser events to server logs. It is a filter you apply, not a link you click.
Status
Section titled “Status”Signup is open at app.getlumin.dev/signup. Every account starts on the free Hobby tier; the plans page lists the paid tiers.
Logs, host + container metrics, container log tailing, Prometheus /metrics scraping, session timelines, uptime checks, and Slack alerting all work end-to-end.
Not built yet: session replay, any clickable pivot between a session and its logs, and alerts that fire from a failed uptime check — alert rules watch metrics and log patterns only. Funnels exist but are gated to instance admins while they are finished.