I Built a Tool That Tells You If It's Worth Paddling Out

I asked a website whether it was worth paddling out at Hilo Bay this morning, and it gave me a real answer — actual wave height, actual water temperature, pulled live off a NOAA buoy sitting in the bay right now, in realtime -- not a cached number from an hour ago. You can go try it yourself: lukelenentine.com/tools/surf-check.

I built it as a test of something bigger than surf reports. I directed the build and tested every piece of it myself, but the actual code came out of a session with Claude Code. That's kinda the real point. I wanted to see if I could drive that process all the way to something with an MCP server so that I could chat with it. Next step is the phone app. But I'm sharing this so others who might feel intimidated by AI can see that it's not that difficult.

The first version took about twenty minutes. A page that fetches a public buoy feed, reads off the numbers, prints them. No AI in it anywhere. Useful in the same way a weather app is useful, and about as interesting to have built.

Once I gave it the actual MCP server, wired up against the public API that already existed, I was suddenly able to chat with it.

That's the whole shift. I was at a job expo today (thanks Star Advertiser!) and I was surprised at how many conversations I had with people that could make real immediate use of AI in their workplaces but were too intimidated to begin.

Security concerns was another issue I heard come up quite a bit. Let's break that into two categories:

1. Approvals - It's true you do need to get the approvals of your IT department, your legal department, and any leadership roles where you'll need their buy-in (or where they've had bad experiences with AI, in which case you'll REALLY want their buy-in). But those are just conversations that can happen once you hand them the prototype product to play with (note: keep FERPA in mind, don't use real personal identifiable data if you're in an institution that is prohibited from doing so, take the time to think through how to achieve your goals while side-stepping PII, you can do it!)
2. Technical Implementation - Again, here I would defer you back to your AI of choice. They're your partner here and can walk you through all the technical implementation details.

In the case of my surf check tool, I added two separate limits on how often anyone can ask it anything: one per visitor, and a second, independent one for the whole site, for the whole day, regardless of who's asking. I even set the second one tighter than it strictly needed to be. This will help protect your wallet since real money is on the line.

The raw tool isn't wide open either. It asks for a key of its own before it'll talk to anyone. I can hand that out to someone who wants to see it work up close.

The version live right now is a website with a chat box on it. The version that would actually be finished watches conditions on its own and tells you something changed before you think to ask — the same shift, one step further, from something you check to something that checks on you. I haven't built that yet. I know exactly what it would take, and it's a longer piece of work than an afternoon, which is exactly why it isn't done.

What I have instead is proof that the middle step is real: a tool that exists, that is secure, and that can be a model to point to when helping others adopt AI solutions.