Developmentversion v1.0.0

Random Data Generator

Uniform integers, floats, booleans and hex bytes for test fixtures and mock data — or a line picked from a list, or the whole list shuffled. Every value comes from a CSPRNG, drawn by rejecting out-of-range bytes rather than folding them, so each outcome really is equally likely.

$RANDOM is a 15-bit generator and $((RANDOM % 6)) is measurably biased; shuf is absent on macOS and jot on Linux. This URL answers the same question from anywhere curl runs.

Not for credentials

For a machine secret in a .env file use /secret. For a password a person has to type use /password.

Both are on the same alphabet-and-entropy footing this tool deliberately does not claim. Nothing here is weaker — there is no seed and the source is the same CSPRNG — but a fixture generator is not where a credential should come from.

Draw

Rangeinclusive at both ends

Count1–1000

Drawn0 values

drawn in your browser — never in this page's HTML

Outcomes

101outcomes

Interval: [0, 100] inclusive
Each: probability exactly 1/101

Drawn by masking random bytes and rejecting the ones that land outside the range — never by folding them with %, which would make the low end of the range more likely.

Why there is no seed

?seed= is a 400, not an ignored parameter. A seeded stream over HTTP would freeze this implementation into a permanent public API, and a test that only reruns green while a website is up has not been made deterministic.

For reproducible fixtures, seed a generator inside your own process, where the feature actually works: random.Random(42) in Python, new Random(42) in Java, rand::rngs::StdRng::seed_from_u64(42) in Rust, faker.seed(42) with Faker.

Privacy

Drawn locally in your browser. The list you paste is never sent to the server.

No database, no logs, no analytics, no URL, no local storage — refreshing discards everything. The command line necessarily draws server-side and the value crosses the network under TLS, exactly as it does for /secret.

From the command line

The same URL answers curl with the values alone — one per line, nothing to strip. A list goes in on stdin with --data-binary @-, and comes back one item per line.

$curl -fsS 'https://tools.primeforge.app/random?type=int&min=0&max=100'

$seq 1 49 | curl -fsS --data-binary @- 'https://tools.primeforge.app/random?type=shuffle' | head -6

$printf 'alice\nbob\ncarol' | curl -fsS --data-binary @- 'https://tools.primeforge.app/random'

Sampling without replacement is a shuffle piped through head -n K — which is precisely what shuf -n K is.