How to analyze confidential data without uploading it

Most handy data tools are hosted. Using one on customer records, payroll, or patient data is a disclosure decision, whether it feels like one or not. Running the analysis in your browser removes that trade-off. You keep the convenience, and the file stays on your machine.

What "uploaded" really means

Once a hosted tool has your file, it sits on a server you don’t control. It’s probably replicated, often kept after your session ends, and readable by the vendor’s staff and by anyone who breaks in. Deleting it later doesn’t undo the transfer.

For data under GDPR, HIPAA, or a plain confidentiality clause, the upload is what creates the obligation. The analysis is the easy part.

How the browser avoids it

WebAssembly lets a real database engine run inside a browser tab. NoEgress ships DuckDB built for WebAssembly. When you open a file, the browser reads it from disk into the tab’s memory and the query runs right there.

Put simply, the engine comes to the data. The app has no upload endpoint, so there’s no server copy to secure, keep, or delete.

Don’t take our word for it

Privacy claims should be easy to check. Open your browser’s developer tools, go to the Network tab, and open a file. You’ll see requests for the app code and the WebAssembly engine. You won’t see your data.

For a stronger test, load the page, turn off Wi-Fi, then open a file and run some queries. Everything keeps working, because nothing in the query path needs a server.

Your browser enforces it too. Every NoEgress page is served with a Content-Security-Policy that tells the browser to refuse connections to any other site, and the “0 sent” badge in the workspace lists every request the page made, with a receipt you can download.

What about the AI assistant?

The optional assistant runs a small open-weight model on your GPU through WebGPU. The weights download once and stay in your browser cache. After that, your prompts, schema, and results are processed on your device and never sent to a model provider.

You don’t need it at all. SQL, pivots, profiling, and charts all work without loading a model.

Where your data is stored

Opened datasets are cached in the browser’s own storage (OPFS), so reloading the page doesn’t make you open the file again. That storage lives on your machine, in your browser profile. Clear the site data and it’s gone. Nothing syncs anywhere.

When this is the right tool

It fits exploratory work on sensitive files: reviewing an export before you send it, checking a vendor report, profiling customer data, answering a one-off payroll question. For scheduled pipelines and shared dashboards you still want a proper warehouse. They solve different problems.