Logs, traces and metrics in one place you run yourself. Point any OpenTelemetry exporter at it. Data lands as compressed Parquet in your own bucket.
Every signal arrives over OTLP and is stored the same way, so a log links to its trace and a dashboard can mix all three.
Check your endpoints from where your users are, and see every host, pod and cloud resource with its health, from the data you already send.
A batch is acknowledged only once the JetStream buffer has stored it. Workers write ZSTD Parquet to object storage, and queriers read it back with a warm cache.
New, spiking and dropping patterns compared with an earlier period, with alerts on new error patterns.
See the logs around any result from the same source, to find what led up to an error.
Jump from a log line to its trace. Kubernetes rollout markers show on the histogram.
Build dashboards, or bring over a Grafana or OpenObserve dashboard's panels and layout.
Error logs, failed spans and their shared traces grouped into incident sessions.
Rules over PromQL, log searches or trace latency, sent to webhooks, Slack, PagerDuty or email.
Claude Code usage per project: cost, tokens, cache use and sessions.
Claude and Cursor search logs, inspect traces and run PromQL with your own access.
Pick Nginx, Postgres, Kafka, Kubernetes and more; get a ready-to-run Collector config.
Retention per project and signal, tail sampling that keeps whole traces, log head sampling.
Who changed what and when. Keys show last use, and unused keys are easy to clean up.
Group monitors in folders. A child of a down parent shows Impacted and stays quiet.
Run several probes per location. If one stops, the others take its monitors within a minute.
Load balancers, databases, buckets and functions from AWS and Azure join your hosts and clusters.
Slack, PagerDuty, webhook and email channels defined once and shared by every project.
Prometheus metrics on every role, a ready Grafana dashboard and SLO alert rules.
Start small. The same image grows into a highly available deployment when your volume does.
Installs Docker if needed, starts Obzly with Postgres and Valkey, and restarts on reboot. About two minutes.
Quick install →Copy the example compose and .env, set a Postgres password, run docker compose up -d.
Three brokers, plus api, worker and querier replicas on Kubernetes, with Parquet in your bucket.
Helm chart →One control plane in front of regions, environments or business units. Data stays in each Site.
Control plane & Sites →No paid tier holds these back. They ship in the same Apache 2.0 image.
Entra ID, Google Workspace, Cognito and others. IdP groups map to project and Site roles.
Users and groups from Okta or Entra ID into organizations and teams. Deprovisioning deactivates.
Required for admins. One switch in Settings requires it for everyone, and every reset lands in the audit log.
Central identities and per-Site access grants. A Site's query API never faces the internet.
Backup, restore, DR and upgrade runbooks, a measured restore drill and obzly reconcile.
Ingest rate, storage, retention and project limits per organization, enforced on every Site.
Your data stays on your infrastructure, and you or your security team can audit all of the code.