Skip to content
QuickStatus.dev

Checks

Add a check

Checks are ongoing probes from the console. They are not the homepage website checker. A component is a row on a public status page — you can link a check to one if you want incidents or metrics to follow the probe.

After you add a check, the detail page covers the bar chart, results, pause, and Run now. The same checks are available from the API.

Check types

Checks in the sidebar opens one list of every check at /app/checks. Filter it by type or status, or search by name or URL. Old per-type links such as /app/checks/api, /app/checks/website, /app/checks/dns, and /app/checks/tcp open the list filtered to that type.

Type What “up” means by default
HTTP (website or API) Any HTTP status below 500 — or, if you choose a code under Up when, Exact HTTP status 200 (or the code you pick)
Heartbeat Your job pinged its URL within the expected period plus grace. See cron job monitoring for crontab, GitHub Actions, Kubernetes, and backup examples
Server Enough servers are reporting through the server agent, and no Mark down rule is firing
MCP The MCP handshake works and tools/list returns at least one tool (and every tool you list)
DNS The lookup succeeds (and expected records match, if you set them)
TCP The connection succeeds — or fails, if you chose Connection should not succeed

Watching an LLM provider? Monitor the OpenAI API walks through an HTTP check step by step: POST body, auth header, and a JSONPath assertion.

On the free plan you can have 10 checks, a minimum interval of 5 minutes, and 1 probe region. The list shows {used} of {max} checks on the {plan} plan. Samples on free are kept for 14 days.

Add a check

  1. Open Checks and click Add check (/app/checks/new). With no checks yet, the list shows the same choices.
  2. Pick a type: Website or API, Heartbeat, Server, MCP server, DNS, or TCP port.
  3. Fill in the Check section. Everything else is folded and has sensible defaults.
  4. Click Add check. The new check's page opens, and the first run starts right away.

Check

The thing being checked comes first. Name is optional except for heartbeats: leave it blank and the name comes from the URL, domain, or host:port.

HTTP

Field Notes
URL Required. Placeholder https://example.com. If you type example.com with no scheme, https:// is added on blur and on submit.
Up when Any response below 500 (the site answered) — the default, so a 403 from bot protection still counts as up. Or Status is 200 OK (or another code) for an exact match, which suits APIs.
Name Optional.
Check every 5, 15, 30 minutes or every hour on Free (every minute and every 2 minutes on Pro).

Save rejects private, loopback, and non-http(s) targets.

Heartbeat

Field Notes
Name Placeholder Nightly backup.
Expected every How often the job runs: 1 minute to 1 day, not below the plan minimum.
Grace Extra wait before a late ping counts as missed. Defaults to the same as Expected every; 0 means none.

After you save, the ping URL is shown once. Call it when the job finishes (curl -fsS -m 10 --retry 3 <ping URL>), and add /fail to report a failed run. Lost it? The check page has Lost it? Get a new ping URL; the old URL stops working.

Server

Field Notes
Name Placeholder Production web servers.
Servers that must report Default 1. The check goes down when fewer servers report.
Grace before down How long it may stay below that number first. Default 5 minutes.

After you save, the check page shows the install command once. Run it on each Linux server. Rules, install options, and fleets are on Server agent.

MCP

Field Notes
MCP server URL Required. Placeholder https://example.com/mcp.
Name, Check every As for HTTP.

Authorization (folded) adds request headers such as Authorization: Bearer …. Expected tools (folded) lists tool names that must be present. Tools are listed, never called.

DNS

Field Notes
Domain Hostname only — example.com, not a URL. Trailing dots are stripped.
Record type A, AAAA, CNAME, TXT, MX, NS, or CAA. Default A.
Name, Check every As for HTTP. Blank name = the domain.

Lookup (folded): Protocol UDP or TCP (default UDP), and an optional DNS server. Leave the server blank to use the probe’s system resolver.

Private, loopback, and .local / localhost names are rejected. To see the current answers from 16 locations, use the DNS propagation checker.

TCP

Field Notes
Hostname Hostname or public IP. No scheme or path.
Port 1–65535. Required.
Name, Check every As for HTTP. Blank name = host:port.

Connection (folded): Expected outcome — Connection should succeed (default) or Connection should not succeed — IP version (IPv4 or IPv6, default IPv4), and an optional Payload sent after connect.

Private, loopback, and link-local targets are rejected.

HTTP request

Folded on create. HTTP only. Summary shows the method.

  • HTTP method — GET, POST, PUT, PATCH, DELETE, HEAD
  • Optional HTTP authorization username / password
  • HTTP body — disabled for GET and HEAD. Hint: Don't forget to add the correct Content-Type request header below.
  • Request headers — add or remove rows
  • Follow redirects — on by default

Running on Kubernetes? See Kubernetes ingress monitoring from outside for which hosts, DNS names and ports to check.

Regions

Folded on create. Summary: 1 of 5 regions on Free, or All 5 regions on Pro. Each box is {name} ({region}), for example San Francisco (us-west-1). The plan’s region cap is selected by default (San Francisco when it is in the list). You must leave at least one selected.

These five regions are not the 16 cities on the homepage checker. Enterprise can add regions on request.

Location Region
San Francisco us-west-1
Ashburn us-east-1
Montreal ca-central-1
Frankfurt eu-central-1
Singapore ap-southeast-1

Each selected location gets its own probe on your interval.

Assertions

Folded on create, all optional. Summary is None until you add one.

HTTP

  • Text to search for (optional) — exact, case-sensitive match in the response body
  • Extra assertions: Body contains, JSONPath equals, Status equals

DNS

  • Expected record values — add one or more values the lookup must include
  • Leave them empty to treat any successful lookup as up

TCP

  • Response contains — optional text that must appear after connect (useful with a payload)
  • Connect under ms — optional maximum connect time

When a 200 still means down

When to alert

Folded on create. Every type except heartbeat and server. Hint: Getting alerts while the site is fine? Raise these.

  • Mark down after — 1 failed run, or 2 / 3 / 5 failed runs in a row (default 3). Why one failed check is noise
  • Mark up again after — 1 / 2 / 3 successful runs
  • Request timeout (HTTP, MCP, DNS) or Connection timeout (TCP) — 5 / 10 / 15 / 30 seconds

HTTPS

Folded on create. HTTP and MCP.

  • Verify the certificate — on by default; the check fails if the certificate is invalid or expired
  • Warn before the certificate expires — Off, 7, 14 (default), or 30 days before. Read once a day; one alert at each step (for example 14, 7, and 1 day). Uptime is not affected. https URLs only: changing a check's URL to http:// turns the warning off and forgets the stored certificate, and changing it back to https:// starts it again at 14 days.

Slow responses

Folded on create. HTTP and MCP. Summary is Off until you set it.

  • Slow above (ms) and Measured on (Total time or Time to first byte). A run that works but is slower marks the check Slow, not down. You get an alert, and a linked component shows Degraded performance. Uses the Mark down after and Mark up again after counts.

Alerts

Folded on create. Send email alerts to your account email. Telegram chats (every plan) and Slack, Discord, PagerDuty, Pushover, and webhook destinations (Pro) are created on Settings → Alerts and attached here. A new check starts with the Telegram chats that get alerts for new checks; see Telegram alerts.

Alerts section on a check: email alerts to the account address, plus an Ops webhook and a #acme-incidents Slack destination.

A check that stays down sends one down alert and, when it recovers, one recovery alert. Email and every attached destination follow that together. A notice that is no longer true when it is sent is dropped: a down alert after the check has recovered, and a recovery alert after it has gone down again. If it goes down again inside the quiet window after that recovery, you get one message that the check is flapping, and further down and up alerts pause. Notices already waiting from the earlier cycle are dropped too. The quiet window is at least 30 minutes. A check on a slower interval waits longer, so two failures of a 5-minute check are still one flap. When the check then stays up, or stays down, through that window, you get one last message (back up, or still down) and normal alerts resume. The checks list still shows each down and up.

Status page

Edit check form on the detail page for Public API, with Your check and HTTP request open.

Folded on create. Hint: Link this check to a component on any of your status pages. Summary starts as Not linked.

  1. Pick a Status page, or leave None.
  2. Pick a Component. Options already linked to another check say already linked and cannot be chosen. One check per component.
  3. Pick what happens under When the check is down: Don't create an incident (default), Create an incident automatically, or Ask me to review first. It stays disabled until you pick a component.
  4. Optionally check Show this metric on the public status page.

If you have no pages yet: Create a status page first to link this check.

With Create an incident automatically, the first confirmed failure opens one incident and keeps it open across a short recovery. With Ask me to review first, the failure puts a draft under Incidents → Needs review instead: Alerts include a link to review the draft. Nothing is published until someone approves it under Incidents.

With either option, an Advanced fold appears. Resolve the incident automatically when the check recovers is on by default. Turn it off and The incident stays open until someone resolves it. See Incidents.

You can also create and edit checks with an API key.