Nevel

Nevel landed at the start of my story with Claude Code, in 2026. Years of domain knowledge and practice met a new tool, the work sped up about twenty times, and my daily routine went out the window xD I designed the first Nevel back in 2019, and if you've read the other stories, you know: now I've come back and built it a whole new system, end to end.

Nevel keeps its books in 1C. The owner is sure there's nothing better than 1C and has only ever worked in 1C. I never did grow to love it ;) It's built for everyone at once, which makes it incredibly tangled, and the code in it, by the way, is written in Russian. But I wasn't going to break what the owner loves.

First we needed a bot with access to 1C. Our 1C developer said honestly that he couldn't do much. That's actually how my hands-on work with Claude Code began, right inside 1C: an HTTP service appeared there, and Telegram was hooked up to it. The owner was happy, and here's why. Nevel has two showrooms and a site with strong SEO, but the site and 1C weren't connected at all, so to find out a price everyone called the owner, and he looked it up on his computer or in Excel. Now the price and the stock come in Telegram.

Next we agreed to start with the store and connected it to 1C. The hardest part turned out to be not that but the move: the old site ranked well in search, and losing those positions wasn't an option. We sat down with 301 and 302 redirects and carried over every address. The new nevel.uz launched on 13 August 2026, a store with a strong foundation and strong SEO. Rankings didn't drop, they went up. The numbers are a little further down this page :)

Then we took on the call center. Nevel had one, but there had been no call analytics since 2022: a call came in, they talked, they sorted it out, and that was it. We wrote our own softphone in Flutter, and it's one of the hardest pieces of work of my life. It took me a month to realize I'd gotten into something very complicated: SIP refused to work the way I needed. Regular softphones don't take a call when the app is closed, and the customer hears a busy tone. For a store that gets calls all day, that means lost orders. We added native parts in Swift and Kotlin, and the call started getting through. The customer is pulled up from the database, and a regular customer lands with the seller they've already talked to. Now we're moving on to call analysis with AI.

The owner asked me not to give too much away, so this is about 10% of what we did xD