Telemetry Machine

npx add-mcp 'https://telemetrymachine.com/api/mcp'

Answer any question about your OTel traces, logs, and metrics.

Traces

FILTERS

Spans

Statusstatus_code

Service resource.service.name

Duration duration_ns

14:0214:0814:1414:2014:2614:32

41 TRACES 1 ERRORS 1240ms P95

41 traces · ready

STATUS TIME ↓ SERVICE SPAN NAME DURATION VS SLOWEST SPANS

ok frontend POST /api/checkout 182ms 12

ok product-catalog GET /products 24.6ms 3

error checkout POST /checkout 286ms 9

ok cart GET /cart 12.8ms 4

ok frontend GET /api/products 38.2ms 6

ok payment POST /charge 89.4ms 3

ok checkout POST /checkout 156ms 11

ok product-catalog GET /products/:id 850ms 2

unset frontend GET /api/cart 1.24s 5

Want to play with real data?

Go see our demo.

Get paged less. Find the cause faster.

Fewer pages, and a head start on every one you do get.

Soon

Get paged only when users are affected.

SLOs track what your users actually feel, and alert on real error budget burn instead of every blip.

checkout availability99.9% over 7d

38% of error budget left

Burning 2.9× too fast. Out in about 2 days.

See what's different about the failures.

Select the slow requests on a heatmap and Surface shows the version, region, or customer that sets them apart.

app.version = 4.2.0 81% vs 12%

cloud.region = eu-west-2 64% vs 23%

http.route = /checkout 47% vs 19%

Soon

Wake up to a write-up.

When an alert fires, an agent queries your traces, logs, and metrics, runs Surface, and checks what changed.

02:14checkout availability SLO alert

Ran Surface on the failing spans

Checked deploys from the last 6 hours

Read error logs from checkout

Likely cause: app.version 4.2.0 in eu-west-2 Rolled out 38 minutes before the burn began.

Dump your proprietary agent. We are OpenTelemetry native.

Send directly from OTel SDKs or through your own Collector.

Instrument with abandon.

Our hyper-efficient engine changes what you can afford to observe. Give your engineers and agents the whole story.

Bring all your context.

Production is only part of the picture. Kube events, change data capture, CI/CD logs, and dev environments help you see what is going on beyond mere prod traces.

Kube eventsKube stateChange Data CaptureCI/CD logsDev environments

Cardinality without anxiety.

Add customer IDs, pod IDs, git SHAs, and any other attribute that helps you find the answer.

customer.id customer_8f2a 48,210 values

k8s.pod.uid checkout-7d9c 3,904 values

git.sha a3f8e21 1,288 values

request.id req_92b4 96.1M values

Your agent cantake it from here.

Whether your service is already instrumented or you're starting from scratch, a single prompt gets you set up.

Our agent skills mean you never have to instrument your services by hand again.

Read the setup guide

Agent session

›

Connect this application to
Telemetry Machine.

Instrumentation addedOTel

Dataset and key createdTM

Production data verifiedOK

**Your first traces are in.**Follow progress in onboarding.

1. 1
2. 2

   Connect your agent

   Add our MCP server to Claude Code, Cursor or Codex.
3. 3

   Let your agent do the rest

   It instruments your app, creates datasets and keys, and works until your traces and logs appear.

Pricing that won't jump-scare you.

One flat monthly price. Spend it on events, metrics, or any mix of the two.

• Starter

  $50 a month 50 million events / 250 million metric points (mix and match) Events $1.00 / million Metric points $0.20 / million

• Team

  $200 a month 250 million events / 1.25 billion metric points (mix and match) Events $0.80 / million Metric points $0.16 / million

• Business

  $500 a month 750 million events / 3.75 billion metric points (mix and match)

• Scale

  $1,000 a month 2 billion events / 10 billion metric points (mix and match)

Every plan includes SLOs, 60 days of retention and unlimited seats.

See full pricing

We believe all software benefits from observability.

That's why we made it dead simple to set up and actually affordable.

14-day free trial. No credit card required.