OpenTelemetry-native observability

Every signal in one place. A bill you can cap.

Logs, metrics and traces linked from the alert to the request, release and log line behind it, an AI SRE that shows its evidence, and the CO₂e of every service. Priced only by the data you send, with a monthly cap that can keep a sample instead of surprising you.

Free: 10 GB of ingest a month and 7-day retention. No card to start.

Open-standard OTLP in and out, so your data stays portable and works with the tools you already run.
LMTLMT
Built on open-standard OpenTelemetry, so your data stays portable Alerts checked every minute Data isolated per tenant, stored in Europe

One pane of glass

From a signal in any layer to a root cause

Logs, metrics and traces from the browser down to the database land in one place, get linked to each other, and reach an AI SRE that shows its evidence and asks before it acts.

Illustrative visualisation with example values, not live data.

Correlation

One session, assembled from every layer

A log line opens the request that wrote it. Errors carry a sample trace. Releases sit on the same charts as latency and 5xx rate, so you follow one story from the browser down to the database instead of five tabs.

Linked signalsTrace ↔ log ↔ metricOne investigation

Linked, not fragmented

Signals that belong to the same request stay connected, so opening an alert shows the related logs, traces and release without a hunt.

Context when it matters

When an incident opens, the AI SRE starts from the assembled picture, not an isolated alert, and lists its evidence and what it ruled out.

logstracesmetrics Frontend AI & LLM API Application Database Network Infrastructure Ingest gateway Linked store Alerts Incidents SLOs AI SRE

Illustrative visualisation with example values. Hover, focus or tap a layer to see what it covers.

The platform

Everything you run, in one place

Pick an area to see how it works. Each animation is an illustration with example values.

Find the line, then the request that wrote it

Search log lines, see volume by level, and open the trace behind any line. Logs, traces and metrics stay linked.

level:error service:checkout
Explore Logs →

Why Owlpane

Four reasons teams switch

One OpenTelemetry-native product where signals link to each other — with limits stated plainly on every comparison page.

01 · Forecast

A bill you can forecast

Priced by GB ingested, not per host, per seat or per indexed event. Starter includes 100 GB; a spend cap samples instead of surprising you at month end.

Cost calculator →
02 · Linked

Logs, traces and metrics that meet

A log line opens the request that wrote it. Errors carry a sample trace. Releases sit on the same charts as latency and 5xx rate — so you follow one story, not five tabs.

See it working →
03 · Standard

OpenTelemetry in and out

OTLP and Prometheus remote write. Send a copy beside Datadog, Grafana or your self-hosted stack, and keep your data portable.

Integrations →
04 · Operations

Kubernetes, SLOs and carbon on the same telemetry

Cluster inventory with pod-level drill-down, burn-rate SLO alerts, and CO₂e per service and job — measured or estimated, always labelled.

Owlpane Carbon →

Compare honestly: Grafana stack · Bronto · Logz.io · Datadog · Checkly

Documentation

Guides, SDKs and ingest setup

Quickstart, OpenTelemetry ingest, Node.js and Kubernetes agent install — the same paths linked from the console setup wizard.

Browse documentation →
Works with

OpenTelemetry, Node.js, NestJS, Kubernetes, PostgreSQL, Redis, Prometheus, Grafana, Slack, PagerDuty, Opsgenie, Jira, and ServiceNow each have a documented connect path. See all integrations →

See it working

The real product, not a mock-up.

Every screen below is a screenshot of the Owlpane console. Watch what each one tells you, or click a step to jump to it.

Know how your API is doing the moment you open it.

Traffic, errors and latency for every back-end service, the map of who calls whom, and the errors grouped by what went wrong.

  1. Overview
  2. Service map
  3. Errors
Explore Applications →
Owlpane · Applications · Overview
Owlpane Applications overview: request, error and latency tiles, request and response-time charts, and a table of routes.

The problem

You're paying five vendors to answer one question

"What just broke, and why?" shouldn't take five browser tabs and a Slack thread to answer.

01

A different tool for every signal

APM here, log search there, uptime somewhere else, on-call in another tab. Nothing shares a trace ID.

02

Pricing that punishes growth

Per-host and per-seat bills that go up exactly when your product is finally getting traction.

03

A cluster you can bill but not see

You know what Kubernetes costs. You don't know why a pod OOMKilled at 2am until a customer tells you.

04

Alerts nobody trusts anymore

Too noisy to page on, too quiet to catch the thing that mattered. So everyone mutes the channel.

Investigate

Follow the incident, not the tabs.

A deploy, a latency jump, an error spike, a slow query and a backed-up queue are one story. In Owlpane each of them already links to the exact requests, errors and logs behind it.

  • Releases are marked on the charts, so you see what changed and when
  • Every chart and card opens the matching requests, errors, logs or traces
  • Alerts carry the evidence that fired them, and resolve when it clears

PreviewIllustrative incident. Each signal exists in the product today on its own page, linked to the requests behind it; the Incident timeline page puts deploys, alerts, uptime results and markers on one axis.

checkout-api · production
  1. 02:57checkout-api v2.14.0 deployedDeploy
  2. 03:01p95 latency 86 ms → 412 msLatency
  3. 03:025xx rate 0.04% → 6.8% · new error PaymentTimeoutErrors
  4. 03:03SELECT customers · p95 3× slowerDatabase
  5. 03:04job SyncInvoices · 12 failed runsJobs
  6. 03:04Alert fired · high 5xx rate · on-call pagedAlert
  7. 03:07v2.13.9 redeployed (rollback)Deploy
  8. 03:09p95 back to 91 ms · alert resolvedRecovered

Setup

Three lines. No config file to fight with.

One call, before your app boots. It does nothing until you point it at an ingest URL — so it's safe to leave in code that hasn't gone live yet.

  • Auto-instruments HTTP, your database driver, and outgoing calls
  • Traces NestJS @Cron and BullMQ jobs with one install line, and any other job through owlpane.job()
  • Tags traces with the signed-in user (an opaque id you pass with owlpane.setUser()), for per-customer debugging
import { owlpane } from "@owlpane/node";

owlpane.start({
  service: "payments-api",
  release: process.env.GIT_SHA,
});

// that's it — every request, query,
// and outgoing call is now traced.

Ask, don't hunt

Ask your telemetry a question.

Plain-language search, saved investigations and a read-only MCP server for your own assistants. They answer and link to the evidence; they never change anything in your systems.

Ask

The charts under each answer are the same aggregates as the rest of the console, so you can check the answer instead of trusting it.

Investigations

Save notes, console links and search URLs for an incident or a root-cause write-up, in one board your team can reopen.

Read-only MCP

Let a coding assistant query topology, flows, incidents and carbon. Read-only by design: there is no auto-remediation.

LLM and agent usage

Calls, tokens, latency and cost from OpenTelemetry gen_ai spans. Prompt and completion text are not read or shown.

See AI & MCP →

How it works

From your code to your dashboard

Your apps

Spans, logs and metrics over standard OTLP.

Node SDKOTel SDKsK8s agent
Ingest gateway

Key resolves the tenant. Secrets are redacted when telemetry is shown or exported.

rate limitsquotasampling
Isolated storage

Per-tenant rules enforced in the database. Designed for an EU region; the live cell is not confirmed yet.

row policiesretention
Console

Drill from any chart to the requests behind it.

alertsSLOsuptime
01

Instrument

Drop in the SDK, or point any existing OpenTelemetry exporter at Owlpane's ingest URL. Works with the OpenTelemetry you already run.

02

Ingest

Telemetry lands in a tenant-isolated cell. Tokens and secrets are redacted when telemetry is shown or exported.

03

Understand

Dashboards, traces, and alerts are there immediately — no query language to learn, no dashboard to build first.

04

Attribute carbon

The carbon layer joins the same telemetry with node energy and hourly grid intensity to attribute CO₂e per service, endpoint and job.

What you need

A free workspace and an ingest key from Setup. Then either @owlpane/node, any OpenTelemetry SDK or Collector, or the Kubernetes agent.

What is redacted

Auth headers, cookies, query strings, emails, bearer tokens, card-like numbers and password or token values are redacted when telemetry is shown or exported. This site does not claim they are removed before storage. The Kubernetes agent has no permission to read Secrets, and the console shows ConfigMap key names only.

If you send more than your plan

Free pauses new ingest above 10 GB a month until you upgrade to Starter. Starter and Business honor the spend cap you set by sampling, so the bill can't jump.

Where it lives, and for how long

Designed for an EU region, isolated per tenant in the database. The query window is 7 days on Free, and 30 days for traces and logs and 90 days for metrics on Starter and Business. Those figures are not physical deletion unless the lake retention sweep is turned on. Standard OTLP out if you ever leave.

Owlpane Carbon · on every plan

Energy and CO₂ are engineering signals too

Monitoring shows latency, errors and cost. Nothing tells a developer that this endpoint or this job made the software use more energy. Owlpane Carbon adds exactly that — on the telemetry you already send.

checkoutmeasured searchestimated nightly-jobmeasured gatewayestimated authmeasured

Illustration with example values. In the console every figure shows whether it was measured or estimated, with its uncertainty.

Per service, endpoint and job

CO₂e per request and per run, following the ISO-standardised Software Carbon Intensity method (ISO/IEC 21031:2024). Version 1 is duration-weighted — the bias is stated on the figure, not hidden.

Carbon regression alerts

Each service is compared with the previous equal window. When its energy or CO₂e rises past your threshold (20% by default) an alert fires, like an error regression. A window comparison, not a per-deploy attribution.

Waste separated from work

Reserved-but-idle capacity reported as its own figure, with the workloads responsible — energy you pay for that does no work.

Measured or estimated, always labelled

Hardware counters where they exist, published factors where they don't. Mode, factor-set version and uncertainty band on every number.

Carbon-aware timing

Grid intensity by hour (hourly for EU zones where ENTSO-E data is available) shows when the grid is cleaner. Hints only — Owlpane does not schedule your workloads.

Telemetry diet

How much telemetry you keep and the storage energy it implies — monitoring's own footprint, as an order-of-magnitude estimate rather than a measurement.

The method: SCI = (E × I + M) per R — energy times grid intensity, plus embodied hardware, per request. Grid intensity from hourly ENTSO-E when configured, with a versioned static factor table as fallback. The methodology is published and linked from the console Factors view.

Available on every plan, Free included, in the product console (/observability/carbon). Metered hardware validation on your own cluster is recommended before operational targets — see the published methodology. The Carbon product page →

Switching

Already on something else? Keep it. Try Owlpane beside it.

Because ingest is standard OpenTelemetry, you can send a copy of your telemetry to Owlpane without touching your current tool. Here is what that looks like, including what you would give up.

Coming from

Datadog

Keep

Any OpenTelemetry instrumentation you already have. Your Datadog agent and account can stay in place while you compare.

Gain

One bill priced by data, with a cap that samples instead of surprising you. No per-host or per-module line items.

Give up

Breadth. Datadog has a much larger integration catalogue, plus network and security products we do not offer. Apps on the proprietary tracer need OpenTelemetry first.

Coming from

Sentry

Keep

Sentry for what it does best, if you like. Send your backend traces and logs to Owlpane next to it.

Gain

Errors grouped by the release that introduced them, with a sample trace, beside your database calls, jobs, Kubernetes state and uptime checks.

Give up

Depth on the front end. Session replay and mobile crash reporting are thinner here, and in beta.

Coming from

Grafana, Loki, Tempo

Keep

Prometheus remote write works today, and Grafana can read Owlpane metrics through a Prometheus-compatible endpoint.

Gain

One managed product instead of four to operate, with logs, traces and metrics that link to each other.

Give up

Ecosystem. PromQL support is a read endpoint, not full parity, and there is no library of community dashboards.

Datadog, New Relic, Grafana, Grafana Cloud, Loki, Tempo, Dash0, Better Stack, Bronto, Logz.io, Amazon CloudWatch, AWS, Sentry, Dynatrace, Kubeshark, Checkly, PagerDuty, SigNoz, Kubernetes, OpenTelemetry, Prometheus and other names are trademarks of their respective owners. Owlpane is not affiliated with or endorsed by them. Comparisons reflect publicly available information as of September 2026 and our own product today, and may change.

Send a copy. Change nothing else.

Add one exporter to your OpenTelemetry Collector. Your current tool keeps working, and Owlpane receives the same data.

Start with one service or ten percent of traffic, compare the answers you get during a real incident, then decide.

collector.yaml
exporters:
  otlphttp/owlpane:
    endpoint: https://ingest.example.com
    headers:
      Authorization: "Bearer ${env:OWLPANE_INGEST_KEY}"

service:
  pipelines:
    traces:
      exporters: [your_current_exporter, otlphttp/owlpane]

Pricing

Priced by data, not by how many people you hire

Unlimited users, projects, hosts, clusters and alert channels on every plan — shared platform, EU data region. You pay for data, and nothing else.

Storage
Billing
Currency

Prices exclude VAT and other taxes. Billed in the currency you choose.

Free
$0forever
Side projects and trying it out for real. No card, no expiry.
  • 10 GB ingested per month
  • Ingest pauses above 10 GB until you upgrade to Starter
  • 7-day query window for traces, logs and metrics
  • Unlimited projects and teammates
  • Every alert channel, unlimited Kubernetes clusters and hosts
  • Carbon: energy and CO₂e per service, endpoint and job
Get started free
Starter
$29/ month
100 GB included, then a per-GB meter.
  • 100 GB included, then $0.35/GB ($0.25/GB past 1 TB)
  • A spend cap you set: your bill can't surprise you
  • 30-day query window for traces and logs, 90-day query window for metrics
  • Unlimited projects
  • Email, Slack, PagerDuty, Opsgenie, Jira and ServiceNow alerts
  • Sampled above 2 TB a month, stopped at 4 TB
Get started
Business
$88/ month
Same 100 GB as Starter. +$59/mo is SSO, SCIM and audit, not more data.
  • Same 100 GB ladder as Starter, then $0.35/GB
  • $29 plan + $59 for SSO, SCIM and audit
  • SSO (OIDC, SAML) and SCIM, audit log export
  • 30-day query window for traces and logs, 90-day query window for metrics
Get started on Free or Starter
Enterprise
Custom
Everything in Business, plus private ingest, isolated storage, and procurement terms — we provision with you.
  • Everything in Business
  • Private ingest endpoint and isolated storage
  • Retention set to your contract
  • Security review and contract terms, including a DPA once counsel adopts one
  • Named contact — sales contract, we provision with you
Contact sales

Lake keeps 30 days. Realtime keeps 7 days: the plan base plus an access fee of $20/mo, including 100 GB, then $0.22/GB ($0.16 past 1 TB). Lake+Realtime is both of those bills. A 14-day window doubles the realtime charge. Realtime does not keep history past the hot window.

Estimate your bill

Free is up to 10 GB a month. Starter is $29/mo including 100 GB, then $0.35/GB ($0.25 past 1 TB). Business is that same 100 GB plus $59/mo for SSO, SCIM and audit ($88/mo). The storage mode follows the selector above.

$76.95/ month
about 6.2 GB per dayStarter estimate

You are billed for what the ingest gateway keeps, not for what is stored afterwards, so later retention does not change the meter. Past 2 TB kept in a month, ingest is sampled and only the kept share is billed; at 4 TB kept, new ingest stops until the month resets.

Questions

Straight answers

Is Owlpane built on OpenTelemetry?
Standard OTLP end to end. Our SDK is a thin, opinionated wrapper over the official OpenTelemetry libraries — anything that already speaks OTLP can send to Owlpane today, and everything Owlpane sends can be read by any other OTel backend if you ever leave. No format lock-in. The Node SDK samples production traffic at 10% by default, and you set the rate.
Where is my data stored?
In Europe. The service runs on Hetzner in Germany, account data sits in a database in Frankfurt (AWS eu-central-1), and telemetry is stored in Cloudflare R2 in Western Europe. Traffic reaches us through Cloudflare's global network. The sub-processors page lists each provider. Tenant data is separated by per-tenant row policies in the telemetry store, not just filtered in application code. Published retention is the query window for your plan, not a promise that stored objects are deleted on that schedule.
What counts as a gigabyte on the bill?
The meter is the size of the telemetry requests our ingest gateway keeps, as your SDK or collector sent them: compressed, as received at the gateway. Requests sampled over the monthly allowance are not counted. Traces, logs and metrics are counted the same way. A gigabyte here is 1,073,741,824 bytes. There is no charge per seat or per host, and retention never changes the meter.
What does a Starter bill look like?
Starter Lake is $29 a month and includes 100 GB. If you send 150 GB in a month, the 50 GB above the allowance cost 50 × $0.35 = $17.50, so the bill is $29 + $17.50 = $46.50. Beyond 1,024 GB the extra gigabytes cost $0.25 each instead of $0.35. Business Lake adds a fixed org fee on the same ladder ($59/mo USD), so Business Lake is $88 at that included volume. Realtime starts at $49 on Starter and $108 on Business. Lake+Realtime starts at $78 on Starter and $137 on Business. History past the hot window is not kept on Realtime. These are USD figures before VAT; EUR and GBP prices are set separately and shown by the currency toggle in the pricing section, and the estimator there does this sum for any volume.
What happens if I send more than my plan includes?
On Free, new ingest pauses above 10 GB a month until you upgrade to Starter or the month resets. Data already stored keeps its retention window. On Starter and Business you can set a monthly spend cap in the console (at least the base fee) and choose what happens when it is reached: keep a 10% sample and pay only for what is kept, or pause ingest until next month. Without a cap, paid plans keep accepting data and bill usage; ingest is sampled above 2 TB a month and stopped at 4 TB.
Do you support Kubernetes?
Application-level tracing works today for any service in a cluster. The Kubernetes agent (node and pod health, OOMKills, rollout status via a minimal Helm chart) is in beta — install from the console setup guide.
Is there a browser / RUM SDK?
Yes, in beta: the @owlpane/browser SDK sends Core Web Vitals, errors and fetch spans that link to server traces when you propagate trace headers.
What languages and frameworks are supported?
Node.js and NestJS have a first-class SDK today (@owlpane/node), with tracing for NestJS @Cron and BullMQ jobs and an owlpane.job() wrapper for other background work. A browser SDK (@owlpane/browser) is in beta. For Java, Go, Python, Ruby and anything else, use the standard OpenTelemetry SDK pointed at your ingest URL — it works with Owlpane right now; thin Owlpane wrappers for Python, Go, Java and Ruby are still in progress.
How is this different from Datadog or Dynatrace?
Both are mature, broad platforms. Owlpane is built on the open OpenTelemetry standard so your data stays portable, prices by data volume rather than hosts or seats, and is built for teams who want one tool instead of five.
Can we self-host it?
Enterprise includes private ingest, isolated storage, and residency options we provision with you. Free, Starter, and Business run on our shared platform (designed for an EU data region). Owlpane is not a download you run yourself.
Is carbon live?
Yes, on every plan, Free included. Open Impact → Carbon in the console once your agent or SDK sends node metrics and server spans. How far back you can look follows your plan's retention. Every figure shows whether it was measured or estimated, with its uncertainty. We still recommend validating against your own meters before using the numbers as hard targets.
Which workloads will carbon cover at general availability?
Kubernetes, VM/Docker hosts and serverless/FaaS — the same surfaces the agent and SDKs already observe. Estimated energy on virtual machines is less accurate than measured energy on bare metal, and every figure will show which mode produced it.
How accurate is per-endpoint carbon?
Version 1 attributes a service's energy across its requests by duration, not per-span CPU time — endpoints that mostly wait on I/O are over-credited, and the UI says so. Request totals come from an unsampled counter, never from sampled traces. The methodology, factor tables and validation results are published, and every figure carries an uncertainty band.

Ready when you are

Stop stitching together five dashboards at 3am.

Create a free workspace in minutes. Need Business, Enterprise, yearly billing or help onboarding Carbon on a large fleet? Email hello@owlpane.com and tell us about your stack.