I built 18 browser tools that keep your inputs local

#privacy #showdev #typescript #webdev

I kept running into the same small technical tasks: formatting JSON, decoding a JWT, calculating an IPv4 subnet, checking a CRC, or converting storage units.

There are already websites for most of these jobs, but I often hesitated before pasting data into an unfamiliar page. I could not always tell whether the input stayed in my browser, was sent to a server, or ended up in a log somewhere.

That became the starting point for Tools in a Tab.

The main constraint: keep tool inputs local

Tools in a Tab is a collection of 18 free browser-based utilities for data, networking, and electronics.

The important distinction is that the values entered into the tools are processed locally in the browser by default. They are not sent to a Tools in a Tab backend for processing.

There is also:

No account requirement
No database handling tool inputs
No SPA
No global UI framework
No server-side processing in the current catalogue

This approach does not fit every possible tool, but it works well for deterministic transformations and calculations.

What is currently included

The data and text group contains tools for:

JSON formatting and validation
JSON, YAML, and CSV conversion
Base64 and URL encoding
JWT decoding
UUID generation
SHA-256 hashing
Text comparison

The networking group includes IPv4 subnet calculations, CIDR and mask conversion, bandwidth calculations, and IPv4 number representations.

The electronics group currently includes CRC calculations, LED resistor sizing, resistor color codes, and storage unit conversion.

Each tool also has an individual static URL, explanatory content, related resources, and reviewed examples.

Architecture

The site is built with Astro and strict TypeScript and deployed as static assets on Cloudflare Workers.

Astro generates the HTML during the build. Interactive JavaScript is loaded only by pages that need it, and the calculation logic is kept separate from the interface.

I deliberately avoided adding a backend or database because neither is necessary for the current tools. Apart from reducing infrastructure, this makes the privacy promise easier to understand and verify.

The site is also available in English and Spanish, with separate canonical URLs and language alternatives.

What was harder than expected

The interface was not the most difficult part. The edge cases were.

A few examples:

Preserving large JSON integers without silently losing precision
Distinguishing valid JWT decoding from signature verification
Handling IPv4 /31 and /32 networks correctly
Supporting multiple CRC models without pretending CRC is one universal algorithm
Distinguishing decimal GB from binary GiB
Keeping every result keyboard-accessible on mobile

Writing tests for known vectors and boundary cases ended up being as important as building the controls.