Blog / Build with AI /

A launch checklist for a small tool

You finish the tool on a Thursday night. It works on your laptop. You are tired, a bit proud, and you want to post the link everywhere before you lose your nerve.

Don’t. Not yet. Give it an hour with a checklist first.

Most small launches don’t fail because the idea was bad. They fail because the link was broken on mobile, the contact page didn’t exist, or the first person to try it hit an error and never came back. Those are boring problems, and boring problems are cheap to fix before launch and expensive after.

Here is the list I’d run through for any small tool, whether it’s a web page, a Chrome extension or a little script with a front end. Most of it takes minutes. Together it’s the difference between “launched” and “shared a half-finished thing”.

Does it do the one thing?

Write one sentence: “This tool helps [who] to [do what].” If you can’t, the launch will be confusing.

Then try it as a stranger would:

  • Open it in a private window, with no saved settings and no logged-in state.
  • Do the main job from start to finish without any help from you.
  • Give it a deliberately bad input: an empty field, a huge file, an emoji, a number where text should be.

If the bad input produces a blank screen or a raw error message, fix that now. A friendly “That didn’t work, try a smaller file” is enough. If you built the tool with an AI assistant, this step matters twice as much, because generated code often handles the happy path beautifully and ignores everything else. I wrote a short list of mistakes to check before you trust AI-written code that is worth a re-read here.

Test it where people will use it

You probably built it on a big screen in your favourite browser. Your visitors won’t all be there.

  • Phone. Open it on a real phone, not only the browser’s device mode. Check that buttons are tappable and nothing runs off the edge.
  • Another browser. If you live in Chrome, try Safari or Firefox.
  • A slow connection. Throttle it, or use mobile data on a bad day. In a lot of places, data isn’t cheap and signal isn’t steady, so a tool that loads six megabytes of scripts for one button will lose people.
  • Keyboard only. Can you tab through it? It is a quick accessibility check that catches a surprising amount.

The security basics

Small tools get away with a lot, but a few things are non-negotiable:

If your tool calls a paid API, set a spending cap on the account before launch. A tool that goes a little bit viral can cost real money overnight.

The pages people look for

Visitors, and the stores or directories you may submit to, expect a few things to exist:

Page or item Why it matters
A clear homepage or landing page One sentence, one screenshot, one button
A privacy note Say what you collect, even if the answer is “nothing”
A contact method An email address or form that someone actually reads
A version number So you and users can say which one is which
A way to report bugs Even a simple form or email

If it’s a browser extension, the store will also want a privacy policy and a justification for every permission you ask for. I covered the writing side in privacy policies for small extensions.

Make the first minute easy

People decide quickly. The first sixty seconds should get them to a result, not to a settings screen. I wrote more about that in onboarding for small tools: the first sixty seconds. For the checklist, three questions are enough:

  1. What does a new visitor see first?
  2. How many clicks until they get something useful?
  3. If they get stuck, what do they see?

If the answer to the third is “nothing”, add an example, a sample input or a one-line hint.

Measure something, lightly

You want to know whether anyone used it, but you don’t need to track people around the internet. A simple, privacy-friendly count of visits and button clicks is enough to learn from. If you aren’t sure how to set that up, adding analytics to a site you built with AI walks through it.

Decide in advance what “it went okay” looks like. Ten real users? Five pieces of feedback? Naming the number now stops you moving the goalposts later.

Prepare the words

Launching is half building and half telling. Before you post anything, write:

  • A one-line description you can paste anywhere.
  • Two or three screenshots, or a short screen recording. Screenshots and demo videos for a tool covers what works.
  • A short “why I built this” paragraph. Honest and specific beats polished. Say what annoyed you and what it does now.
  • A list of places to share it. Not twenty. Three or four where the right people actually hang out.

If you want advice on finding the first people, getting your first users for a tool is the natural next read.

Plan for the day after

The launch isn’t the end. Before you press the button:

  • Put a backup of the current working version somewhere safe. If you aren’t using version control yet, git basics for non-developers is a gentle start.
  • Decide when you’ll check feedback (say, morning and evening for three days) so it doesn’t take over your life.
  • Have a rollback plan: if something breaks, can you put the old version back in five minutes?
  • Keep a list of requests. Don’t promise anything on launch day that you’d resent building in a month.

Support is also part of launching. If it’s just you, read supporting users when you are a team of one before the first email arrives.

The short version

Copy this somewhere and tick it off:

Key takeaways

  • Boring failures sink more launches than bad ideas.
  • Test as a stranger, on a phone, with bad input.
  • Have the privacy note, contact method and version number ready.
  • Measure lightly and decide what success looks like before you launch.
  • Plan the day after: backup, rollback and feedback time.

What’s the one item on this list you’d be most tempted to skip? That’s probably the one to do first.