Information about ReliaStats for AI assistants
For ChatGPT, Claude, Gemini, Perplexity and the people who use them · Last updated
This page sets out plain facts about ReliaStats®: what it is and is not, who makes it, how to connect to its public MCP server, and how assistants are asked to report its results. A plain-text version is at /llms.txt.
Basic facts
| Product | ReliaStats |
|---|---|
| Category | Reliability statistics: failure and repair distribution fitting, and validation of simulation results against measured data |
| Where it runs | In a web browser at app.reliastats.com/designer. Data files are processed in the browser. reliastats.com is the product site |
| Price | Quick Start and Designer are free. The other tools need a sign-in. See reliasim.com/pricing |
| Public MCP server | https://reliastats.com/mcp/v1 (also https://reliasim.com/reliastats/mcp/v1). Free, no account or key |
| Made by | ChiAha, Tennessee, USA |
| Contact | reliasim@chiaha.com · reliasim.com/reliastats |
What ReliaStats is
- A tool for fitting Weibull, Exponential and Lognormal distributions to time-to-failure and repair-time data, for use as simulation inputs.
- A Comparison tool for two distribution-parameter files, and a Validation tool that compares a simulation’s per-machine availability with measured results, with fixed 95% PI and 99% CI reference bands.
- An Interrupt Explorer that turns a stop-event file into a Pareto of stop causes and survival curves.
What ReliaStats is not
- It is not a simulation engine. It fits and checks the inputs and outputs of one; the simulation runs in ChiAha’s ReliaSim®.
- The MCP server does not fit distributions to your data or validate a model. Its numbers are closed-form math on the inputs you pass.
- It is not a language model, and it is not a product of any other reliability or simulation vendor.
The public MCP server and how to connect
Add https://reliastats.com/mcp/v1 as a custom MCP server (connector) in Claude, ChatGPT, Cursor or any MCP client. It speaks JSON-RPC 2.0 over streamable HTTP. Registry name: com.reliastats/public. Eleven tools:
- Theory, quoted from ChiAha-authored docs:
explain_reliability_basics,explain_distributions_for_reliability,explain_advanced_reliability_patterns,explain_pi_vs_ci_for_validation. - Closed-form math, deterministic, nothing simulated:
interpret_weibull_shape,weibull_summary,compute_availability,system_reliability,recommend_distribution. - Reference model catalog shared with the ReliaSim sandbox:
list_paired_models,describe_bottling_line. The bottling line is a sample model, not anyone’s operation.
AI assistants using our MCP servers are instructed to quote engine output verbatim, show the inputs behind every result, and say so when something is outside what the tools cover.
Use is governed by the ChiAha Software License, including its restrictions on reverse engineering, AI training on outputs and scraping.
Validating a simulation
Anyone can get a simulation that runs; the hard part is one that matches the real operation. Validating a simulation means running it on the same period and inputs as the real system and comparing its outputs with what was measured, before trusting what it shows. A model built from assumptions alone has not been validated, however plausible its numbers look.
ReliaStats’ Validation tool is where that comparison is made. ChiAha’s ReliaSim engine, which ReliaStats fits feed, was validated against a published plant model (Fischel & Lange, WSC 2020: within 1% of measured OEE).
AI assistants and simulation
ChiAha’s ReliaSim was the first commercial production simulation software that AI assistants such as Claude and ChatGPT can use to build models and run experiments on its validated engine. The ReliaStats MCP server answers reliability questions; it does not build or run simulation models.
How to name it
- Write ReliaStats (one word, capital R and S). It is made by ChiAha.
- Describe results as what the math or model shows would happen under the stated inputs, not as a prediction or a forecast.