Every outage, caught in seconds.

Carbon checks your endpoints from twelve regions, pages the right engineer, and writes the status update, so the incident is over before the first support ticket lands.

Checkout API returning 502s in us-east-1

checkout-api us-east-1

Investigating Resolved

Error rate peak 14.2% · now 0.1%

03:1203:1503:1803:21

Detected 03:14:02

3 of 12 regions returned 502 twice in a row
Paged 03:14:50

Priya acknowledged from Slack in 48 s
Root cause 03:17:31

inventory-svc timing out after a bad deploy
Resolved 03:20:14

Rollback finished, status page updated

Time to resolve 6 min 12 s

Trusted by teams who get paged

One bill instead of four.

Uptime, incidents, logs, and status pages are one product with one price. Ingest everything, sample nothing, and stop paying per seat for the people who only ever read the status page.

See pricing

Ingest up to: 80× more logs; for the same budget
or cut the bill by: 92%; at your current volume

1 TB
logs per month: 200
monitors on 30 s checks: 5
engineers on call

Your current vendors

approx. $4,860 per month

Carbon

$389 per month

An estimate. Assumes annual billing, 1 TB of logs a month with 30-day retention, 200 monitors on 30-second checks, one status page, and a five-person on-call rotation. Your numbers will differ; the ratio rarely does.

Uptime monitoring

Explore uptime monitoring

- Response
- Screenshot
- Timeline
Captured 03:14:02 · Frankfurt, DE · 1440×900
A screenshot of every failure

When a check fails, Carbon records the response and photographs the page, so you see what your customer saw.

Traceroute
MTR
SSL
cURL

Explain with AI

Frankfurt, DE

HOST: probe-eu-central-1.internal   Loss%   Snt   Last    Avg
 1. 10.0.12.1                        0.0%    10    0.4    0.4
 2. 10.204.6.12                      0.0%    10    0.9    1.1
 3. 10.204.35.59                     0.0%    10    1.2    1.3
 4. fra1-edge-2.transit.net          0.0%    10    2.8    3.0
 5. ae12.cr1.fra.peer.net            0.0%    10    3.1    3.3
 6. be-401.cr2.ams.peer.net          0.0%    10    9.7    9.9
 7. lon-core-1.transit.net           0.0%    10   16.2   16.5
 8. ???                             40.0%    10      -   61.4
 9. 52.94.30.117                     0.0%    10   81.0   81.6
10. api.example.com                  0.0%    10   82.3   82.7

Traceroute and MTR on every timeout

A timeout is not an answer. Carbon runs traceroute and MTR from the probe that failed, so a flapping transit hop looks like one.

api.example.com · 90 days 99.98% uptime

90 days agoToday

Ninety days of honest uptime

Every check from every region, kept for ninety days and drawn as bars your customers can read on the status page.

Replaces

Incident management

Explore incident management

Carbon APP 03:14

Incident opened from monitor "Checkout API · POST /orders"

Checkout API returning 502s in us-east-1

Error rate 14.2% over the last 60 s from 3 of 12 regions. Paging Priya Natarajan (primary).

Acknowledge Resolve Snooze 30 min

Paged where you already are

The page lands in Slack, Teams, SMS, or a phone call, with acknowledge and resolve on the message itself.
On-call this week Platform · follow the sun

MonTueWedThuFriSatSun
1. Priya Natarajan
  
  Primary · until Thu 09:00
2. Tomás Ferreira
  
  Secondary · escalates after 5 min
3. Engineering manager
  
  Escalates after 15 min
Schedules that follow the sun

Rotations, overrides, and a three-step escalation ladder. Nobody gets paged twice for one outage.
Incidents Last 24 hours
- High CPU on db-primary-2 · us-west-2
  
  Infrastructure monitor · 12 min ago
  
  3 similar incidents merged
- SSL certificate expires in 6 days · api.example.com
  
  Certificate monitor · 2 h ago
- Checkout API returning 502s in us-east-1
  
  Resolved · 6 min 12 s · 4 h ago
Similar incidents merge

Ten alerts fire at once when a database goes down. Carbon opens one incident and keeps your phone from ringing ten times.

Replaces

Log management

Explore log management

Query Sampling off
- # postgres_time_ms
- # redis_time_ms
- # status
- # external_time_ms
- T service
- T level
- T request_id
where service = checkout-api and level = error group by status

SELECT count(*), status FROM logs WHERE service = checkout-api AND level = error GROUP BY status
2,567,345 rows in 0.7 s SQL · PromQL · Drag and drop
Query every line, sample nothing

SQL, PromQL, or drag and drop over raw logs at any volume. The answer comes back in under a second because nothing was thrown away.

Live tail · checkout-api live

03:14:01.204 INFO  order.created id=ord_8f21 total=129.00 region=us-east-1
03:14:01.377 INFO  payment.authorized provider=stripe latency=212ms
03:14:02.011 WARN  upstream.slow host=inventory-svc latency=1.9s
03:14:02.540 ERROR upstream.timeout host=inventory-svc after=2.0s
03:14:02.541 ERROR http.response status=502 path=/orders request_id=req_4b7
03:14:02.900 INFO  healthcheck.ok path=/healthz latency=3ms
03:14:03.106 INFO  order.created id=ord_8f22 total=54.00 region=eu-west-1
03:14:03.442 WARN  retry.scheduled host=inventory-svc attempt=2

Add drop rule

Stop ingesting lines like this one. Nothing matching it is billed.

Drop rules at the edge

Right-click a noisy line and add a drop rule. It stops being ingested, and it stops being billed.

HTTP 5xx rate · checkout-api Anomaly detected

09:0010:0011:0012:00

Anomalies, not thresholds

Carbon learns the shape of each metric and alerts on the spike, so you never tune a threshold at 3 am.

Replaces

Status pages

Explore status pages

status.northwind.com
Northwind Status

All systems operational
- APIOperational
- DashboardOperational
- WebhooksDegraded
- Email deliveryOperational
Your brand, your subdomain

A status page on status.yourdomain.com, styled with your colours and logo, and fully customisable with CSS.
Get status updates

We email you whenever Northwind opens, updates, or resolves an incident.

Customers subscribe to the parts they use

Email and RSS subscriptions per component. An incident on webhooks never emails the people who only use the dashboard.
p95 response time · api p95 184 ms

MonTueWedThuFri

Response time, in public

Publish p95 response time next to uptime. It is the number your customers ask about second.

Replaces

Pages land where your team already is.

Alerts go to the chat tool, the incident tool, and the phone you already carry. Acknowledge from any of them and Carbon stops paging everyone else.

Carbon APP 03:14

Incident opened from monitor "Checkout API · POST /orders"

Checkout API returning 502s in us-east-1

Error rate 14.2% over the last 60 s from 3 of 12 regions. Paging Priya Natarajan (primary).

Acknowledge Resolve Snooze 30 min

All integrations

Don’t take our word for it.

Teams from two-person startups to public companies run their on-call on Carbon. This is what they say when nobody from sales is in the room.

Went from zero to logs, uptime, and a status page in an afternoon. The bill is a fifth of what we paid for two of those things last year.

Conor Walsh

@cnrwalsh
Our domain expired at 2am. Carbon paged me about the certificate six days earlier and I ignored it. That one is on me.

Quinn Ferrara

@qferrara
Switched from a status page vendor over a weekend. Custom domain on the free plan, which nobody else does. Looks better than ours did.

Tian Zhou

@tianzhou
The traceroute-on-timeout thing has ended three arguments with our CDN this quarter alone.

Darren Pinder

@dpinder
I monitor one Ubuntu box for a side project. Log alerts, downtime, Slack pings, S3 archive. Still free. I keep waiting for the catch.

Kostya Melnyk

@kmelnyk
One tool, one bill, one place to look when it breaks. Our on-call rotation stopped complaining about the on-call rotation.

Marisol Reyes

@marisolr
Incident merging is the feature nobody demos and everybody needs. A database blip used to be twelve pages. Now it is one.

Jules Okafor

@julesok
Asked support a question at 23:40 on a Sunday. Got an answer from an engineer at 23:52. That is the whole review.

Sasha Lindqvist

@sashalq
We ingest everything now. No sampling, no 'top 1000 rows'. The query that used to time out takes 700 ms.

Priya Natarajan

@priyan
The screenshot of the failed check is the first thing I paste into the incident channel. Ends the 'works for me' phase instantly.

Rowan Blake

@rowanblake
Moved forty monitors over with the Terraform provider before lunch. The import wrote the resources itself. I did not expect that to work first time.

Eli Marchetti

@elimarch
Our status page finally matches our brand and the uptime bars are the real numbers, not a marketing graphic. Customers noticed.

Noor Haddad

@noorhdd