Privacy

Network engineers handle address plans, device configurations and secrets that must not leave the building. NetKit is built so they do not have to.

The short version

  • No accounts, no uploads, no telemetry. There is no backend that stores your input, and no analytics script on any page.
  • Calculations run in this tab. The kernel is a pure TypeScript library with zero dependencies; it cannot open a socket.
  • Shareable links use the URL fragment. A tool's input lives after the #, which browsers never include in an HTTP request — so a link you paste to a colleague does not leak the address plan to a server.
  • Offline by default. A service worker caches the app, so it keeps working with no network at all — which is the real proof that nothing is being sent.

Which tools contact the network

Every tool declares its network use in code, and the build fails if the declaration and the implementation disagree. Current state:

  • Runs locally22 tools make no request whatsoever.
  • Sends a requestNo tool currently needs the network.

How the claim is enforced

“Local-first” is easy to write on a landing page. Here it is a build invariant, checked by scripts/check-invariants.mjs on every commit:

  • The kernel (@netkit/core) is scanned for network APIs, DOM access and any import outside its own directory. Any hit fails the build.
  • Each tool's source is scanned for fetch, XMLHttpRequest, WebSocket, EventSource and friends. Finding one in a tool that declares network: false fails the build.
  • A tool that declares network: true but makes no request is reported as well, so the badges cannot drift in either direction.

What NetKit cannot promise

Your browser, your extensions and your network are outside our control. If you are on a managed device, assume anything you type could be logged locally. What we can promise is that this code sends nothing anywhere — and you can verify it by loading the page, disconnecting, and using it.

Not engineering advice

The tools are provided to save time and reduce arithmetic mistakes. Generated configuration, in particular, must be validated on a lab device before it reaches production. Where a vendor-specific detail is not confirmed on hardware, the interface says so rather than presenting it as fact.