For years, my wife and I had used food trackers that were either too expensive, too gimmicky, or both. The nagging was relentless. Our own data sat buried behind a paywall. Logging a banana triggered a confetti explosion — as if we’d run a marathon. We wanted something calmer, so my wife and I could track what we ate and actually keep doing it past the second week. The answer came when I started exploring Claude Coding — and what followed genuinely surprised me. For anyone curious about how AI-assisted development works in practice, the NCSC’s guidance for developers is worth a read alongside it.
So I built it. The app is called FoodPulse, and it is the first iOS app CMS Web Solutions has shipped. It lives under Digital Pulse, the brand CMS Web Solutions now uses for its own apps and future products; our client websites carry on here, under CMS Web Solutions. Through Claude Coding, the whole process moved faster than I expected — though not without friction. This is the honest story of how it came together, including the mistake that embarrassed me and the lesson that saved me.
Why we built FoodPulse with Claude Coding

Most tracking apps treat you like a project to be optimised. Streaks, badges, red warnings when you go over — all designed to keep you poking the screen rather than getting on with your life. I wanted the opposite: a tool that respects your attention and your privacy. Claude Coding made it possible to focus on those principles from day one, rather than getting bogged down in boilerplate.
We built FoodPulse around four principles, and they shaped every decision.
- Fast logging. Common actions take one or two taps, and repeating yesterday’s meal should be effortless.
- Honest numbers. Anything the app estimates is clearly labelled as an estimate. No false precision.
- Private by default. Your food logs and weight stay on your iPhone and in your own private iCloud. Our server never sees your diary.
- Calm design. No streaks, no confetti, no nagging. Go over a target, and it shows in amber, never red.
That last point matters more than it sounds. A single night out, or a weekend away, shouldn’t feel like failure. So the app shows a rolling seven-day calorie balance, which means one big day evens out across the week rather than ruining it.
What Claude Coding actually built

At its core, FoodPulse is a food, nutrition and water tracker — but the aim was to make logging genuinely painless.
You can scan a barcode, photograph a nutrition label, or enter food by hand. You can build meals from your own saved foods, and even scan recipes straight from a cookbook. If you can’t be bothered with any of that, you can photograph or describe a meal and let the AI estimate the nutrition for you — labelled as an estimate, of course.
Beyond food, it tracks water, weight and a weight-loss plan with milestones. Exercise calories come in from Apple Health; there’s an Apple Watch app, and everything syncs between your iPhones through iCloud.
A Claude Coding-powered coach that knows what you eat
The feature I use most is the AI Nutrition Coach. It’s a coach you can actually talk to. At teatime I can ask, “I’ve 500 calories left, and I’m short on fibre — what should I eat?” and it suggests meals built from foods I’ve already saved.
It offers general wellbeing guidance, not medical advice, and we’re careful about that distinction. An Android version is in testing alongside the iOS one.
How it was built with Claude Code
Here’s what fellow developers will want, and what small-business owners are probably most curious about. I built FoodPulse with Claude Code, Anthropic’s AI coding assistant, working as a pair programmer.
I made the product decisions: what to build, how it should feel, and — just as importantly — what to leave out. Claude Code wrote much of the code and the automated tests, and checked its own work in the iPhone simulator. I tested every build on my own phones.
The pace of development surprised even me. Work started on 27 October, and within days the ninth test build was on TestFlight for our team of early BETA testers, alongside the Android version, the server and an admin dashboard. That would once have taken months and months of expensive programming time. Those days are over!
Why speed alone isn’t the story
I want to be careful here, because this topic attracts nonsense from both directions. AI did not build this app by itself. It wrote code quickly, but it had no idea what a good food tracker should feel like, where the privacy line should sit, or why amber beats red.
Twenty years of building websites and web software is what made the speed useful. I knew which questions to ask, which suggestions to reject, and when a tidy-looking solution was quietly the wrong one. The AI accelerated the typing; the judgement was still the job.
And no, I’m not frightened for my trade. A faster chisel doesn’t replace the carpenter — it just means you spend more time deciding what to carve!
Behind the scenes, kept simple
The AI features run through our own server rather than on your phone. That’s a deliberate choice: it means the AI key never sits on anyone’s device, and we can keep an eye on how everything behaves.
A dedicated admin dashboard lets us invite testers, set a monthly allowance for each, watch costs, and switch between AI models from a simple drop-down. We can run with Anthropic Sonnet, Opus and Fable. None of this touches your private diary — that stays on your iPhone and in your own iCloud, exactly as promised.
The mistake that caught me out
Now the confession, because no honest build story is all triumph.
iCloud Sync, between my two iPhones, worked flawlessly in testing. I’d log a meal on one phone and watch it appear on the other. Fantastic. Except when I installed the real build on my second phone, it sat there completely empty. Some head-scratching followed.
The cause turned out to be a quirk I’d half-forgotten: Apple keeps separate test and live copies of each app’s iCloud database. My test builds were talking happily to the test copy, but the live copy had never been set up. So test builds were happy, and real installs failed quietly—the worst kind of failure, because nothing shouts at you.
The fix took minutes once we understood it. The lesson took longer to sink in, and it’s one I’d pass to any developer: test the way your users will actually install the app, not just the way developers do. Your clever internal workflow can hide the exact bug your first real customer will hit.
What’s next
Friends and clients are now testing FoodPulse through Apple TestFlight, and we’ll widen that group before any public release. I’m in no rush. A calm app deserves a calm roll-out, and I’d rather it was genuinely good than merely early.
For me, though, the bigger takeaway is what a small studio can do now. A business idea that would once have needed a long, expensive development project can increasingly become a working, testable app in a fraction of the time — provided someone with real experience is steering.
Where to start if you’ve an app idea
If you’ve been sitting on an idea for your business — an internal tool, a customer-facing app, something that would save your team hours a week — the honest first step is a conversation, not a contract. Write down the one job you most want it to do, and the one thing you’d happily leave out. That clarity is worth more than any technology.
Then, if you’d like a second opinion from someone who’s just done it, talk to us at CMS Web Solutions. I’m happy to tell you honestly whether an app is the right answer — even when it isn’t.
Ready to talk?
Let’s start with a conversation.
A 30-minute discovery call — no slides, no pitch. Just a conversation about what you’re working on and whether we’re the right fit.