Integer distributions
random_int
Generate integers from uniform, Bernoulli, binomial, or Poisson distributions.
MCP server · Cloudflare Workers
Reliable randomness for AI agents.
Move random decisions out of the language model. random-mcp generates integer and real-valued distributions, weighted choices, and repeated samples using the Web Crypto API.
Tools
A small, focused toolset
Three tools cover common randomization tasks. Every tool supports batched results through the count field.
Real-valued distributions
random_double
Generate floating-point values from uniform, normal, lognormal, or exponential distributions.
Selection
random_choice
Select one or more strings, with optional weights and sampling with or without replacement.
Why random-mcp
Language models predict. They do not draw lots.
A model may favor familiar-looking values when asked to choose randomly. This server performs the random operation outside the model so agents can rely on an explicit, auditable tool call.
Unbiased integer generation
Integer generation uses rejection sampling to prevent modulo bias, so no value becomes more likely merely because the source range does not divide evenly.
Distributions for each task
Choose uniform, Bernoulli, binomial, or Poisson distributions for integers, and uniform, normal, lognormal, or exponential distributions for real values.
Connect
Add it to an MCP client
Add a custom MCP server
Open your client's MCP connection settings and choose a custom Streamable HTTP server.
Enter the endpoint
Use the MCP endpoint shown above and select OAuth as the authentication method.
Approve and sign in
Approve the requested access, complete GitHub sign-in, and enable the tools you want to use.