Obstack collects traces, metrics and logs without code changes, keeps them on servers you own, and shows them in Grafana. No per-gigabyte meters. No phone-home.
Every product speaks OpenTelemetry, so each one works with what you already run, and the next one slots in without a migration.
The eBPF application agent, on its own.
HTTP, gRPC, TLS, MySQL and Redis become traces without an SDK, and slow Node.js awaits are named down to the function. Bee exports to any OpenTelemetry collector that accepts OTLP over HTTP.
The complete stack, built around Bee.
Bee and Beyla for collection, Vector and Kafka for transport, GreptimeDB for storage, Grafana for dashboards, Keep for alerts and incidents, and the Grafana MCP server for agents. Self-hosted, or run by us in the region you choose. We configure it end to end; custom dashboards, alerts and migration are quoted and billed on time and materials. One support contract.
An agent that works incidents alongside your on-call. It reads the same traces, metrics, logs and alerts your team does, through the stack's own interfaces, on your infrastructure.
Request early accessOur own observability console: an alternative to Grafana, designed around the Obstack data model from the first screen.
Hear when it shipsEvery signal lands in the same store with the same labels, so a slow request, the host it ran on and the device that sent it can be read side by side.
Bee and Beyla read service traffic from the Linux kernel. Requests, database calls and the hops between services become traces with no SDK and no redeploy.
Hosts, containers and Kubernetes nodes report CPU, memory, disk and logs through Vector, while Bee maps which services talk to which.
Gateways and devices send OTLP into the same pipeline, with Kafka absorbing the bursts. Tell us about your fleet at info@obstack.it.
eBPF agents on every node: Bee goes deep on the services that matter, Beyla goes broad on the rest.
Vector gathers OTLP, host metrics and logs. Kafka buffers large or bursty fleets when you need it.
Metrics, logs and traces in one database, queried with SQL and PromQL.
Dashboards and exploration on the Grafana your team already knows. Eagle is on its way.
Keep runs alerting and incidents; the Grafana MCP server lets agents read what your team reads.
Telemetry goes to collectors and databases you run. Licensing works offline too: a signed file, no license server, no beacon. Air-gapped sites are a first-class deployment.
OTLP and Prometheus on the way in, SQL, PromQL and Grafana on the way out. If you ever leave, your data is already in formats every tool reads.
A flat price per node. When an incident triples your traffic, your observability bill stays exactly where it was.
Install the package next to your busiest Node.js service and watch request-level traces arrive in your existing Grafana within the hour.
