We are kofel, a small studio in Tyrol. Everything we build solved our problems first. Maybe it solves yours.
Software that belongs to the device it runs on.
When one codebase serves the product better than two.
Sites that load fast and read well.
Considered tools with state that survives a refresh.
On the device, inside your own data, or agentic with a threshold.
Some work is best kept on the device. Some needs your own data. And some an agent can take over, up to a threshold. Which one is right is the job’s call.
AI that never leaves the phone. A sentence becomes a time, a place and a person, locally, and then a reminder. No cloud call, no data leaving the device.
Not every task needs a frontier model. And where nothing is sent, the question of where the data sits never comes up.
A co-pilot that knows your own data: projects, invoices, notes. It answers from them. Alongside it, receipt capture reads supplier, amount and date off a photo.
An assistant is worth exactly as much as its context. 14 modules, EU hosting, an open beta.
Feedback becomes a draft pull request. Six intake channels, spam and prompt-injection screening, grouping by meaning, and a threshold per project: above it the agent carries on, below it a person.
Autonomous agents do not fail on the model, they fail because nobody said where they stop. Here the customer draws that line.
A long call. We find out what it’s really about before we sketch.
Two pages, plain language. What we will do, what we won’t, where the lines are.
We build in small, visible steps. There is something you can click from early on. You watch it grow and shape it as it does.
We don’t disappear at launch. Updates, fixes, and a real person who answers when something needs attention.
AI takes the drafts, the scaffolds, the edge cases. The judgement, the taste, and the last read of every line stays human.
Two or three people who care is more than ten who don’t.
A premium feel is the bare minimum. You notice the opposite at the first tap.
Most defaults are accidents. Ours aren’t. Each one is a choice we can defend.
Most projects fail in execution because they were never finished in planning. We’d rather plan a little longer and leave less to chance.
We build prototypes and iterate on them before we commit to real code. The cheap version of a mistake is a prototype. The expensive version is a release.
The empty state, the error message, the loading spinner: that is not polish on top of the work. That is the work.