I Asked Claude for a Checklist. I Ended Up Building an App.

I did not set out to build an application. I was trying to keep up with Google Search Console.

I regularly review pages that Google has discovered but not indexed, along with new pages that have not yet been discovered or indexed. I already had an automation keeping the underlying list current. That part was relatively straightforward. The annoying part was the human side of the process.

Which URLs had I already submitted? When did I submit them? Which ones still needed attention? And if I moved from one computer to another, how could I keep that working list in sync?

I started using Claude Artifacts to solve it. What happened next changed the way I think about small, custom business tools.

Google Search Console can tell me what Google knows about a URL. Google also provides a Request Indexing option through URL Inspection for pages I manage. But requesting indexing is not a magic button. Google says a crawl request can take days or weeks and does not guarantee that a URL will be indexed.

That means I still need a process around the process.

I wanted a working queue that helped me answer simple questions quickly: Which pages need attention? Which URLs have I already submitted? When did I submit them? Which ones should I come back to later?

I could have put all of that in a spreadsheet. But I wanted something that behaved more like the way I actually work.

Artifacts can also connect to external services through Model Context Protocol, or MCP, on supported plans. Anthropic specifically documents Artifacts that can read from and write to tools such as Asana, Google Calendar, Slack, and custom MCP servers.

That creates a useful mental model for what an Artifact can become:

Interface: the thing you see and interact with.

State: the information the Artifact remembers.

Connections: the outside tools and systems it can work with.

AI: Claude itself, when the application benefits from analysis, generation, classification, or reasoning.

Put those layers together and you are much closer to a purpose-built business application than a static AI answer.

What else could a small business build this way?

My GSC checklist is one example. Once I understood the pattern, I started seeing other small tools everywhere.

There is a limit to the lesson.

Artifacts are not automatically a replacement for a mature CRM, a transactional database, a secure line-of-business application, or software with complex permissions, compliance requirements, large data volumes, or mission-critical availability requirements.

Even the persistent storage feature has documented limits. And when an Artifact connects to outside systems, permissions and security deserve the same attention they would in any other business workflow.

The opportunity is not to rebuild every piece of software you use. It is to notice the little workflow gaps between those systems.

Those gaps are everywhere.

Start with the annoyance, not the technology

The most important part of this experiment was not knowing how to code an Artifact. It was being able to explain what was irritating me.

I did not begin with a software specification. I began with something closer to: I have this list, I need to find things in it, I need to mark what I handled, and I need those choices to still be there on another computer.

Then I kept improving it through conversation.

Solve that first. Use it. Find the friction. Ask for the next improvement.

That is how my checklist became an app.

This project reinforced something I have been seeing across AI-assisted development: the economics of custom software are changing.

Not every business problem deserves a large development project. But that does not mean the business has to live with a clumsy process forever. There is a growing middle ground where a business owner can describe a very specific need and create a very specific tool around it.

The advantage is not “AI can code.” The advantage is that the distance between an operational problem and a usable first solution is getting much shorter.

For me, this one started with Google Search Console. The more interesting discovery was what else I could release next.

CTA: Have a small workflow that should work better?

Released Solutions helps small businesses turn repetitive work, disconnected tools, and AI ideas into practical systems they can actually use.

Explore AI-Managed Online Presence