Every signal. One platform. Your infrastructure.
Logs, metrics, traces, uptime, alerts and incidents — unified in a single dashboard. Deploy it inside your own network, or let us host it for you.
Then just ask — Copilot answers from your own telemetry, and renders the real view inline.
Your infrastructure, or ours.
ObserveKit is built to run on-premises — the whole platform is five containers and one compose file. If you would rather not operate it, we run the same product for you.
Self-hosted
RecommendedRuns entirely inside your own network, on your hardware or in your cloud account. Telemetry never leaves your perimeter, which takes data residency and third-party processing off the table.
- Server, UI, ClickHouse and object storage — five containers total
- You own retention, access control and the data itself
- Deploys to Kubernetes or plain Docker
- Run the AI Copilot against a local model so nothing is sent out
We host it
ManagedThe same product, operated by Expeed. Point your agents and OpenTelemetry SDKs at our endpoint and start querying — no infrastructure for your team to run or upgrade.
- Identical features — nothing is held back from the self-hosted build
- Upgrades, scaling and backups handled for you
- Start here and move on-prem later; the agents do not change
- Per-source access control keeps teams scoped to their own estate
Two ways in. Both take minutes.
An agent for containers, OpenTelemetry for everything else — including applications installed straight onto a Linux server with no container runtime at all. One store, one dashboard.
Install the agent
One command drops a DaemonSet into your cluster. Zero code changes — you get the infrastructure’s view, including software you didn’t write.
# Kubernetes kubectl apply -f https://<your-observekit>/install/<token>
- Pod & container logs from every node
- Node / pod metrics and cluster state
- TLS certificate expiry, out of the box
Or send OpenTelemetry
Point any OTel SDK at the endpoint with your API key — you get the application’s view. Works from a container, a VM, or a plain Linux box under systemd.
OTEL_EXPORTER_OTLP_ENDPOINT=https://<your-observekit> OTEL_EXPORTER_OTLP_HEADERS=X-API-Key=<key> OTEL_SERVICE_NAME=checkout-api
- Distributed traces & spans
- Custom & business metrics
- App logs, correlated to traces
Every signal, on one screen — and they talk to each other.
A slow trace links to the exact log lines of that request. A failing service links to its metrics. A deploy marker explains both. Stop stitching four tools together in your head.
Follow a single trace_id from a latency spike straight down to the log line that explains it — no copy-pasting between products.
Investigate an incident
Log explorer, live tail, span waterfalls, an automatic service map, and exceptions grouped by fingerprint. Pivot from a span to the log lines it produced.
One tool, one investigationService & infra health
Request rate, error rate and p50/p90/p99 per service, alongside node, pod and container health with Kubernetes state built in.
Is it the app or the box?Alerting & incidents
One expression language over metrics and logs — rates, forecasting, anomalies. Per-entity alerts, routing, silences and escalation policies.
18 ready-made templatesUptime & status pages
Synthetic checks with TLS-expiry tracking, dead-man's-switch monitors for cron jobs, and a public status page you curate for customers.
HTTP · TCP · heartbeatDashboards & reporting
Custom dashboards from counters, charts, tables and log streams, plus ingestion and storage analytics for capacity planning. Works on a phone.
Built by drag-and-dropControl cost and noise
Pipelines extract, mask, rename or drop log lines on the way in — so secrets never land in the store. Hot and cold retention tiers are yours to set.
Filter before you storeAccess & governance
Microsoft Entra ID single sign-on, three roles, and per-source access enforced on the server for every request — not just hidden in the UI.
SSO, roles, source scopingAsk it in plain English
A conversational alternative to every screen. Charts, tables, log streams and trace waterfalls render inline — and every alert gets a written root-cause hint.
CopilotOne problem should wake you once.
A rule is evaluated separately for every pod, container, node or deployment it matches, so each one resolves on its own — but alerts that share the labels you group on collapse into a single incident, and notifications are batched rather than sent one per alert.
Ask anything. Get the real view — not a chatbot’s guess.
Copilot is a first-class alternative to every screen in ObserveKit. Ask in plain language and it answers from your telemetry, rendering the logs, traces, metrics and cost views inline — drawn from the same data, no query syntax or dashboard-hunting.
Everything you can open, you can ask for. Flip between Dashboard and Copilot anytime — same data, two ways to work.
Self-hosting? Point Copilot at a local model and no telemetry ever leaves the building.
Point it at your stack. See everything.
Tell us what you're running and whether you want it in your own network or hosted by us — we'll get you set up with a source and an API key.
Request access
Bring one service or a whole fleet. We'll help you install the agent or wire up OpenTelemetry, and you'll be querying real data the same day.