$ zivvo --list-instruments

seven instruments, one set of checks.

Zivvo does one thing: it tells the right engineer that an API has changed before a customer does. What follows is how each instrument works, what it shows you, and where it sits in a normal week of shipping software.

Get started How it works

01 / capabilities

what each instrument does, and what it is for

Every capability below reads from the same checks. That is deliberate: the page, the diff, the trace and the status history are always telling the same story.

Multi-region checks

one slow zone is a finding, not a rumour

Zivvo calls each endpoint from more than thirty regions and sets every zone against the median of the rest. A single region drifting upward is flagged as the odd one out, with the gap in milliseconds, so “it’s slow for some people” becomes a named location and a number.

  • Intervals from five minutes down to ten seconds, set per endpoint.
  • Latency budgets drawn on every bar, so a breach is visible at a glance.
  • REST and GraphQL calls, with headers and auth held in encrypted variables.

In your weekA CDN change on Tuesday makes one region slower. You see it that afternoon, not in Friday’s support queue.

Schema drift

a field that changes shape arrives with its diff

Zivvo learns the response shape of each endpoint and compares every reply against it. When total_cents turns into an amount object, you are shown the before and after, the deploy it first appeared with, and a plain choice: accept the new shape as correct, or keep alerting until it is fixed.

  • Imports OpenAPI and GraphQL schemas, or learns shape from live traffic.
  • Removed, renamed and retyped fields are separated from harmless additions.
  • Each drift is linked to the commit that shipped it.

In your weekA backend refactor merges at 09:30. By 09:45 the mobile team has the diff and can decide whether the client needs a patch.

Traced alerts

the alert carries its own evidence

When a check fails, Zivvo captures the request: headers, response body and a span waterfall showing where the time went. The person who is paged opens a single link and begins at the cause (here, a ledger database timing out) rather than at the symptom.

  • Trace snapshot, response body and failing zones in one incident drawer.
  • Sensitive headers and fields are masked before anything is stored.
  • Acknowledge, mute or reassign without leaving the drawer.

In your weekA 502 appears in two zones at 09:47. The on-call engineer is looking at the slow database span by 09:49.

Test flows

a journey is tested as a journey

A single ping says little about a checkout. Chain requests into a flow (create a cart, add an item, pay) and assert on status, schema and a time budget at every step. Run it on a schedule, or from CI the moment a deploy completes, so the pipeline can fail on a broken cart call.

  • Values from one step feed the next: ids, tokens, totals.
  • Flows are plain YAML in your repository, reviewed like any other change.
  • A deploy hook runs the flow and returns a pass or fail to CI.

In your weekA merge lands on main. The flow runs against production within a minute and posts its verdict on the pull request.

On-call routing

the page goes to whoever owns the endpoint tonight

Routes map endpoints to rotations. Payments paths page the backend primary; storefront GraphQL pages the web team. Escalate to a secondary after a set quiet period, and mute a noisy endpoint for an hour without silencing the rest.

  • Rotations read from your calendar or from PagerDuty schedules.
  • Quiet hours, severity thresholds and escalation delays per route.
  • Slack, Microsoft Teams, email and PagerDuty as destinations.

In your weekPriya hands the pager to Tomás on Monday at 09:00. Nothing needs reconfiguring; the routes already follow the rota.

Status pages

a status page that cannot disagree with the pager

Public or private, the page is built from the same checks that wake your engineers. Choose which components appear, write the human update on top, and let the thirty-day history show the honest record underneath. It will not stay green while the pager is ringing.

  • Components map to endpoint groups; you decide what customers see.
  • Private pages for staff, behind SSO on the Scale plan.
  • Incident notes are written once and shown on the page and in Slack.

In your weekA customer asks if payments is down. Support sends one link and the answer is already written.

Command palette

every endpoint is two keystrokes away

Press ⌘K and type a path, an incident number or a person. Jump to an endpoint’s history, acknowledge an incident, or mute a route for an hour, with the keyboard doing all of it. During an outage, the fewest clicks matter most.

  • Fuzzy search across endpoints, incidents, flows and routes.
  • Actions share shortcuts with the tables, so the muscle memory carries over.
  • Recent items surface first, so the thing you were just investigating is on top.

In your weekMid-incident, you type “pay” and land on the payments endpoint without touching the sidebar.

02 / security

you hand us credentials, so we treat them like credentials

Monitoring means calling your APIs with real tokens. These are the controls that keep that arrangement boring.

encrypted variables

Tokens and API keys are encrypted at rest, shown masked in the interface, and never written to a trace or an alert message.

private location agents

Check internal services from inside your network with an agent you run yourself. Only results leave your firewall, never credentials.

SSO and audit logs

Sign in through your identity provider, and see who changed a route, accepted a schema or muted an alert, and when.

03 / integrations

it speaks to the tools you already keep open

A ledger of where Zivvo plugs in, and what it does once it gets there.

  • Paging

    PagerDuty

    Incidents open with the trace attached and resolve themselves when the check recovers.

  • Chat

    Slack and Microsoft Teams

    Alerts, drift diffs and status updates post to the channel you choose, with acknowledge buttons inside the message.

  • CI and deploys

    deploy hooks

    Trigger a test flow when a deploy finishes and return pass or fail to your pipeline, with the commit hash shown in every incident.

  • Product teams

    Intercom and Segment

    Annotate support threads and event streams with incident windows, so customer reports line up with what actually happened.

04 / reliability

a monitor that is itself hard to take down

A watcher that goes quiet is worse than none. Zivvo checks run from independent regions on separate providers, a failure in one never silences the rest, and a confirmation step across zones keeps a single bad network path from paging anyone at three in the morning.

  • Failures are confirmed from a second zone before an alert is sent.
  • Our own checks are monitored the same way, from outside our infrastructure.
  • Log retention from seven days to a year, depending on plan.

$ zivvo init

seven instruments. one plan that fits how fast you need to know.

The Hobby plan is free; Team adds flows and on-call scheduling; Scale takes the interval down to ten seconds. Every plan carries drift detection and a trace on every alert.

Get started