Know which deploy broke production.
Stacktrace checks every release against the one before it. When something gets slower, it names the change and puts the rollback one click away.
Checkout response time
p95 · last 2 hours
v2.41.0 made checkout 3.5× slower.
- Response time
- 172ms→618ms
- Errors
- 0.2%→2.9%
Likely cause
a91f3c now looks up prices one item at a time.
Plugs into the way 300 software teams already ship
- OpenTelemetry
- GitHub Actions
- Kubernetes
- Vercel
- AWS Lambda
- PagerDuty
- Slack
01 · Deploys
Every deploy gets a verdict within five minutes.
Faster, unchanged or slower. Stacktrace compares the five minutes either side of each release, so you know how it behaved before standup.
| Service | Version | Author | Time | Before · after | Verdict |
|---|---|---|---|---|---|
| checkout-api | v2.41.0 | @mira | 14:02 | Slower · rolled back | |
| search | v8.3.2 | @kofi | 13:47 | Faster · −31% | |
| auth | v5.0.9 | @ana | 13:20 | No change | |
| notifications | v1.18.0 | @li | 12:55 | Watch · +12% | |
| billing-worker | v3.7.1 | @sam | 12:31 | No change |
02 · Traces
The slow step is already highlighted.
Open any spike and Stacktrace shows one slow request, step by step, with the part that changed since the last release marked in red.
03 · Incidents
Postmortems write their own first draft.
Timeline, impact and cause come straight from the incident. Your team adds the judgement and the follow-ups, and the review starts with facts.
Checkout slowed after v2.41.0
Duration
15 min
Slow checkouts
3.1%
Failed payments
0
What happened
Commit a91f3c replaced one batched price lookup with a lookup per cart item. Large carts went over the 400ms target until the rollback at 14:14.
Follow-ups
- Add a query budget to cart tests@mira
- Flag repeated queries in CI@devon
Why Stacktrace
Simple to start. Fair to scale.
Built on open standards, priced so every engineer can have a seat, and designed for the person holding the pager at 3am.
- Setup
- One OpenTelemetry exporter. First deploy check in 10 minutes.
- Data
- Metrics, logs and traces in one place, searchable with SQL.
- Retention
- 13 months of metrics, 30 days of full traces.
- Pricing
- Per GB ingested. Unlimited users and dashboards.
- Security
- SOC 2 Type II, EU and US regions, SSO on every plan.
- Exit
- Export everything to Parquet at any time.
Use cases
One tool for everyone who ships.
01
Platform teams
One health view for every service, with owners pulled from your service catalogue.
02
Product engineers
See how your release behaved in production before standup.
03
On-call
Pages that arrive with the suspect deploy and a rollback attached.
04
Engineering leaders
Change failure rate and time to recover, by team, from real incidents.
Illustrative outcome · Larkspur
The median incident went from 47 minutes to 11.
Over the first two quarters. Most of the time saved came from knowing which deploy to look at.
Before Stacktrace
47 min
- Paged2m
- Finding the deploy28m
- Fix and redeploy17m
On Stacktrace
11 min
- Paged with the verdict1m
- Roll back4m
- Confirm recovery6m
“The page arrives with the deploy that caused it. Half our incidents now end with a rollback before anyone opens a dashboard.”
Connect your first service before lunch.
Free for 14 days on your real traffic. Point an OpenTelemetry exporter at Stacktrace and your next deploy arrives with a verdict.
$ npx stacktrace init ✓ Found Node 22 and Express ✓ Added the OpenTelemetry exporter ✓ Linked deploys from GitHub → First traces arriving