Message Explorer
One place to see, search and fix every message Qonto sends its customers. I built it for Content Design. Today the whole company uses it.
Overview
Qonto sends thousands of messages: product emails, push notifications, SMS and marketing campaigns. Until this year, nobody could see them all in one place.
Message Explorer is one searchable list of all of them. You can read any message in any language, see the translation key behind each line, and send a colleague a link straight to it. I built it from start to finish with Claude Code.
Context
One customer message could live in four different systems. Product emails, push and SMS are set up in the back-end, and you could only see them on engineering preview pages. Their copy lives in Lokalise. Marketing campaigns live in three separate Customer.io workspaces.
My role
I built it, from the first idea to every release. The idea came from something simple: my teammates needed to find emails and had to open several tools to do it. I built the first version around that daily need and kept improving it by asking people what they thought.
The content challenge
A simple question like "what exactly do we tell a client when we suspend their account, and what does it say in Italian?" meant opening several tools or asking an engineer. Most people asked the engineer. Nobody had the full list of what we send.
I started from how people look for a message, not from how the systems store it.
The tool has one list for every source, which you can filter by source, squad or translation status. You can search by what a message says: paste a sentence from a customer screenshot and the tool finds the message, even when its internal name means nothing to you. Every message can be read in each of the 7 languages, one at a time or side by side. Hover over a line to see its translation key, so you go from "this sentence is wrong" to the key that fixes it in one hover. And you can share the exact message as a link that opens in the right language.
For a content designer: find the message, hover over the line, get the key, fix it. No ticket, no waiting.
How I used AI
I built it with Claude Code. On every release I checked the result myself: does it work, does it hold up as more people use it, does it fit what people need. AI sped up the building. The checking stayed with me.
Outcome
- 5,871
- messages
- 66
- people
- 139
- searches
All in one place, across 6 sources and 7 languages (count from 24 September 2026). In its first 22 days, it was used on all 16 working days, with 663 messages previewed. Usage is counted anonymously: no names, no emails, no IP addresses, and search terms are not recorded.
Search by text, launched on 23 September, made up 31% of that day's opens (60 of 195).
Made for Content Design, now used by CRM, marketing, growth, front end and back end engineers.
Without context you cannot build a tool that is useful to many people.
Starting from a real daily need, and from the people who had it, is what made this one work. What I am proudest of: it was made for Content Design and the whole company now uses it.