Le texte intégral de ce guide n'est actuellement disponible qu'en anglais.
What these tools actually do
Almost every username tool reduces to the same operation: take a string, substitute it into a template URL for each platform, request the page, and classify the response. The differences between tools are in the size of their platform list and the quality of their classifier.
That framing matters because it tells you what to evaluate. A tool with 3,000 sites and a naive "HTTP 200 means found" rule will produce more noise than a tool with 300 sites and a careful one.
Sherlock, Maigret and WhatsMyName
Sherlock is the classic command-line checker: a compact JSON site list, fast parallel requests, plain output. It is easy to read and easy to extend, which is why it is still the default example in tutorials.
Maigret is the larger sibling — thousands of sites, more metadata per site, and richer output formats. WhatsMyName takes a different approach: it is a curated community list of URL patterns and detection rules that other tools consume.
All three share one limitation: they are maintained by volunteers against sites that change constantly. Expect a meaningful fraction of rules to be stale at any moment.
Hosted search versus local tooling
Running a checker locally keeps your queries off someone else's server, which matters if the handle you are looking up is sensitive. The trade-off is maintenance: you update the rules, you debug the failures, you keep the machine online.
A hosted tool removes that work and usually adds presentation — reports, links, exports — at the cost of your queries passing through a third party. For casual self-audits the difference is small; for anything sensitive, prefer local.
Where the software stops and judgement starts
No tool decides whether a hit is the same person. That inference is yours, and it should rest on something the person published — a linked site, a shared photo, a consistent bio — not on the fact that two pages share a string.
The professional habit is to write down what each hit actually showed, separately from what you think it means. The moment those two columns merge, the report stops being evidence.