On-call and incident response
Vesper checks every alert, folds the noise into one incident and runs the steps your team wrote down. It wakes a person only when a step needs one.
Runbooks/Checkout/checkout-api p99 latency
Save draftPublish
Alert: p99 > 1.2s for 3 min
Confirm in 2 of 3 regions
Roll back to v411
Scale to 12 pods
Checkout primary rotation
lgtmDefine what happens next
Keeping nights quiet for
900+ engineering teams
Why Vesper
nice catch!One slow deploy can fire 41 alerts across latency, errors and queue depth, and people learn to swipe them away. Vesper folds them into one incident, confirms it is real and runs the steps your team already agreed on. A page reaches a person only when a person has something to do.
View use cases
INC-2317 at Larkspur: checkout slowed down after a late deploy. Every step on the right ran on its own while the Checkout rotation slept.
Datadog: checkout-api p99 at 1.84s, threshold 1.2s
Confirmed in 3 of 3 regions, not a flap
41 related alerts folded into INC-2317
Deploy in the last 30 min: yes, v412 at 01:52
Rolled back checkout-api from v412 to v411
p99 back to 240ms after 90s
Status page: “Slow checkout” marked resolved
Draft created for Tomás Ortega, for the morning
The steps your best responder takes half asleep, written down once and run in order. Each step logs what it saw and what it did.
Same service, same five minutes, similar signal: one incident. Last month 12,486 alerts became 214 incidents.
Confirmed in 3 of 3 regions · held 3 min
Confirm in more than one region, compare with the last deploy, ignore anything that flaps for under three minutes.
p99 1.84s → 240ms
Roll back, scale out, restart, flush a cache. Every action has a guard, a timeout and a way to undo it.
Rollback did not bring p99 under 400ms. Context attached: 41 alerts, deploy diff, last 3 incidents.
When a step needs a human, Vesper pages whoever holds that service's rotation, with everything the runbook already found.
Larkspur status
The runbook posts the first update within a minute and the resolution when its checks pass. Your customers hear from you before they ask.
Dry-run any runbook against real alerts from the past seven days. See what it would have closed, fixed or paged, without touching production.
Vesper counts the nights, not only the shifts. Night pages are tracked per person, swaps take two taps and handoff notes travel with the pager.
TomásSamMayaHana
Swapped. Sam covers Tue 21, you cover Fri 17.
Vesper imports the timeline, the graphs and the runbook log into a draft. The owner adds what only a person knows: why, and what changes.
INC-2317 · SEV2 · Thu 16 Oct
Larkspur, 40 engineers across four teams, last 30 days on Vesper.
“Thursday nights were the ones people swapped away, because that was release day. Now the runbook takes the first look. Last month the Checkout rotation was woken up twice.”
Point Datadog, Grafana, Prometheus or CloudWatch at one webhook. Grouping starts with the first alert.
Paste the wiki page. Vesper turns the steps into a flow you can edit, then dry-run against last week.
Import schedules from a calendar or another pager. Overrides, swaps and escalation come along.
Start in shadow mode. Switch a runbook to live when its replay matches what your team would have done.
Works with the tools already on your pager
$0up to 5 responders
For a small team with one rotation and a handful of services.
$21per responder / month
For engineering orgs that want runbooks to take the first page.
Customfrom 150 responders
For companies with audit needs and many teams on call.
Still unsure if a runbook should touch production at 3 a.m.? Talk to someone who has carried the pager.
Talk to an engineerEvery action has a guard (for example, only roll back if a deploy happened in the last 30 minutes) and a timeout. If a check fails, the runbook stops and pages the rotation with what it saw. Every action can be undone from the incident.
Yes. Shadow mode runs the whole flow and posts what it would have done in the incident channel. Most teams run shadow mode for one or two weeks, then replay the period before switching a runbook to live.
No. Vesper can sit in front of it and forward only the pages that need a person. Teams that switch fully import schedules and escalation policies in one step.
By service, a time window (five minutes by default) and signal similarity. You can see why each alert joined a group, split it, or write a rule for the next time.
In the EU or US region you choose. Alert payloads are kept for 90 days on Team, longer on Scale. We never train on your data.
Anyone who can be put on a rotation or paged. Viewers, stakeholders and status page subscribers are free.
Free for up to five responders. Shadow mode from day one.
