You have a failing request. You copy the JWT from the Authorization header, search "jwt decoder online", and paste it into the first result. Ten seconds later you know why the token is rejected.
You also just handed a signed token, with user IDs, tenant names and maybe an email address, to a site you know nothing about.
What actually leaks
- JWTs: claims often contain user IDs, emails, roles and internal tenant identifiers. Until it expires, the token may also be replayable.
- JSON payloads: API responses copied from production regularly contain PII, internal hostnames and secrets.
- Configs and logs:
.envsnippets, connection strings, Nginx configs. - Base64 blobs: people forget that Base64 is encoding, not encryption.
A 30-second check for any online tool
- Open DevTools (F12) and go to the Network tab.
- Clear the log, then paste your input and click the tool's button.
- If a request fires with your input in the body or query string, the tool sends your data to a server.
- If nothing fires, processing is happening locally. For extra safety, turn on "Offline" in the Network tab and try again. A client-side tool keeps working.
Note that analytics scripts can load on the page. That is different from your input being transmitted, so look at what the requests contain.
Safer habits
- Prefer tools that run entirely in the browser.
- Use local CLI tools for anything sensitive:
jq,base64,openssl,jwt-cli. - Never paste long-lived production secrets anywhere. Rotate them if you did.
Client-side tools I built for this
I built WebDevToolbox because I wanted these utilities without worrying about where the input goes. They run in your browser, with no signup:
Disclosure: I'm the author. Verify the claim yourself with the Network-tab check above; it works on any tool, including mine.
What developer tool would you like to see added next?

