The night shift

It's 2:13 a.m. Checkout just broke.

Every check still passes. You're asleep. Here are the next three minutes with ItsOpen on the night shift, and every feature in the order it earns its keep.

  1. Every check still says 200 OK

    GET /checkout 200 1.8 s · eu-west, us-east, ap-south
    Uptime checks· From outside, as often as every 30 seconds

    Every URL you care about, requested on a schedule and judged by its status code and content. Two failures in a row open an incident; the first success closes it.

    • Checks from every 30 seconds to every hour, depending on your plan, with timeouts and expected status codes you choose
    • A keyword check catches error pages that still answer 200
    • 90 days of uptime, day by day, and every check of the last month
    StorefrontChecked 4s ago
    99.98%Uptime · 60 days200 OK · 212 msEvery 30 seconds

  2. But this site doesn't answer in 1.8 seconds

    checkout-api p95 1.8 s · usual 210 ms · 8.7× → anomaly
    Anomaly detection· Slow is the new down

    Each site's normal speed is learned from its own history, so a page that usually answers in 200 ms is flagged at 900, and a slow one isn't flagged for being itself.

    • A baseline per monitor that adapts as your site changes
    • Spikes in browser errors judged against your traffic, not a fixed threshold
    • Statistical tests, so a quiet site and a busy one are judged fairly
    Checkout API · response time8.7× usual
    usual

  3. Seven people press Report a problem

    7 visitors /checkout 15 min · p < 0.000001 → incident opened
    Visitor reports· A down detector for your site

    A small script adds a "Report a problem" button to your pages. When several visitors report at once, it's an incident, even if every check passes.

    • One line of HTML: no cookies, no IP addresses, only paths
    • Counts each person once, so one frustrated visitor isn't an outage
    • Your public status page takes reports too
    Visitor reports · 15 min7 people · incident
    • It won't load/checkout1m
    • Something isn't working/checkout3m
    • It won't load/cart4m
    Report a problem

  4. The replay shows a Pay button that does nothing

    session 4f2a rage click ×6 · TypeError checkout.js:212
    Session replay· See what they saw

    Turn on replay and every report links to the moment in the visitor's session: the pages, clicks and errors leading up to it, rebuilt in your dashboard.

    • Our own recorder: what people type and email addresses become dots before they leave the browser
    • Mark any element to mask or leave it out; rage clicks found for you
    • Visitors who ask not to be tracked are never recorded
    /checkout
    Pay now

  5. Customers see it on your status page, not in your inbox

    status/storefront Checkout → degraded · posted automatically
    Status pages· One per project

    Publish a page for each product with 90 days of uptime and current incidents, and let anyone check whether a site is down for everyone or just for them.

    • Choose which monitors each page shows
    • Incidents appear and resolve on their own
    • A public directory and an "is it down?" check for any site
    All systems openUpdated just now
    • Website
    • API
    • Webhooks
    Having trouble? Tell us

  6. Your agent already knows what's wrong, and where

    GET /v1/agent/summary 200 · 1 incident · Checkout API
    Built for agents· API keys and Sign in with ItsOpen

    Give an AI agent read access to one project or all of them, and it can see what's wrong and why in a single call. Apps connect through OAuth with PKCE.

    • An agent API with its own OpenAPI document
    • Scoped keys that end when their maker loses access
    • Incidents can open in the Arcaro Labs hub for its agents to investigate
    $ curl /v1/agent/summary
    {
  "headline": "1 of 12 monitors slow, with 1 open incident.",
  "attention": [{
    "monitorName": "Checkout API",
    "status": "slow",
    "why": "8.7× its usual response time"
  }]
}

02:31 · back open

By morning, it's one incident with the whole story

The slowdown, the reports, the replay and the fix land on the same incident, and it closes itself once the signal is quiet. Nobody stitched five tools together at 2 a.m. Read it with your coffee.

Checkout API slowStorefront · 18 min
Resolved
  1. 02:13Checks pass. The site answers. · Uptime checks
  2. 02:13Checkout is 8.7× slower than usual. · Anomaly detection
  3. 02:14Seven visitors report it. Incident opened. · Visitor reports
  4. 02:15Replay: rage clicks on a dead button. · Session replay
  5. 02:15The status page posts it on its own. · Status pages
  6. 02:16Your agent has the summary, and the line. · Built for agents
  7. 02:31Quiet for ten minutes. Closed itself.

Get started

Keep the sign lit. Hear about it first when it flickers.

Add your first project in two minutes. Free for as long as you like, no card needed.

Prefer email? hello@itsopen.dev