Choose the right monitor type

Choose the right monitor type

Pick HTTP, keyword, TCP, ICMP, or DNS based on the simple question you want Uptime to answer.
9 min read Updated Sep 5, 2026 You use the simplest check that accurately represents customer availability

Begin with one question

Before choosing a monitor type, complete this sentence:

I need to know whether __________.

For example: “I need to know whether customers can open our website.” That question points to an HTTP monitor.

Choose the check that matches your goal

Most websites should start with HTTP. The other options answer more specific questions.

Visual guide

Website or API

HTTP

Best default

Page contains text

Keyword

Content check

Port accepts a connection

TCP

Network service

Host answers ping

ICMP

Only when supported

Domain record is correct

DNS

Resolution check

Quick recommendation

Choose HTTP if you are monitoring a website and are unsure which option to use. It is the safest default for non-technical users.

Your goalChooseExample
Confirm a website or online application opensHTTPhttps://example.com
Confirm a page contains expected wordsKeywordA page contains “Service is available”
Confirm a network service accepts connectionsTCPA mail server accepts port 443 or 25
Confirm a host intentionally answers network pingsICMPA managed router that permits ping
Confirm a domain record resolves correctlyDNSwww.example.com resolves to the expected record

HTTP: websites and APIs

HTTP checks make a normal web request and evaluate the result. Use HTTP for:

  • public websites;
  • online stores;
  • customer dashboards;
  • login pages;
  • public health endpoints;
  • web APIs.

Keep GET, 2xx–3xx, follow redirects, and SSL checks enabled for a typical first monitor.

Homepage or health page?

A homepage proves visitors can reach your public site. A dedicated health page can prove that more of the application is working, but it may require help from your technical team.

Start with the homepage if you do not have a health page. A useful monitor today is better than a theoretically perfect monitor that never gets created.

Keyword: expected page content

A keyword check is an HTTP check with one extra question: does the returned page contain a particular piece of text?

Use stable text such as a product name or a dedicated health value. Avoid prices, promotional headings, dates, inventory counts, or frequently translated content.

If the chosen words change while the site is otherwise healthy, the monitor will fail. Use keyword checks only when the text itself is part of being healthy.

TCP: a service accepts a connection

TCP checks try to connect to a hostname and port. They are useful for services that do not provide a normal web page.

You may need a developer or infrastructure administrator to provide the correct hostname and port. A successful connection proves the service accepted the connection; it does not prove a complete business action succeeded.

ICMP: network ping

ICMP sends a basic network ping. It is appropriate only when the target is designed to answer pings.

Many healthy public websites block ICMP for security or network-management reasons. A blocked ping can look like downtime even while the site works normally. Do not use ICMP as the default website check.

DNS: domain records

DNS connects a name such as www.example.com to the record computers use to find it. DNS monitoring is useful during domain, email, or provider changes.

Common records include:

  • A or AAAA for web destinations;
  • CNAME for an alias;
  • MX for email delivery;
  • TXT for verification and email security settings.

Ask the person who manages your domain which record and expected value matter. A broad resolution check is easier to maintain; an exact value catches unintended changes.

When one service needs more than one monitor

One product can have several customer-important actions. For example, an online store might use:

  1. an HTTP monitor for the storefront;
  2. an HTTP or keyword monitor for checkout health;
  3. a DNS monitor for the public domain;
  4. a separate monitor for its API.

Name each monitor after the capability it represents. During an incident, Checkout is more useful than production-service-03.

Final decision checklist

  • Does this check represent something customers actually need?
  • Is the target publicly reachable from the selected regions?
  • Is HTTP a better fit than ping?
  • Will the expected result remain stable during normal changes?
  • Would a teammate understand the monitor name during an alert?

Next, create the monitor or learn how to read its results.