Summary
Randomness the model cannot fake, split into a fast set and a cryptographic set.
Ask a model for a random number and you get something that looks random and is not; these tools hand the job to random.randint, random.choices and friends. The two secure tools go through Python's secrets module instead — that is the pair to use for tokens and keys, and the plain ones for simulations, samples and games.
What it is
Seven tools over Python's own random and secrets modules — integers, floats, weighted picks, shuffles, samples and cryptographic tokens.
What you get
- Random integers between bounds, inclusive at both ends
- Random floats between bounds, defaulting to 0.0 and 1.0
- A pick of k items from a list with replacement, optionally weighted per item
- A shuffled copy of a list, and a sample of k unique items without replacement
- Hex tokens from Python's secrets module, sized by the number of bytes you ask for
- A cryptographically secure integer below a bound you set
Requirements
Nothing to authorize. Published on PyPI as random-number-mcp and launched over stdio with uvx; Python 3.10 or newer.
Setup effort
One command — uvx random-number-mcp