Conjecture.Mcp

What is it?

Hypothesis's engine, reimagined for idiomatic C#. Write tests that describe what your code should do, and let Conjecture generate hundreds of random inputs to find the edge cases you'd never think of. When it finds a failure, it automatically shrinks the input to the smallest possible counterexample — no hand-written cases required.

Install

Pick the adapter for your test framework:

dotnet add package Conjecture.Xunit      # xUnit v2
dotnet add package Conjecture.Xunit.V3   # xUnit v3
dotnet add package Conjecture.NUnit      # NUnit
dotnet add package Conjecture.MSTest     # MSTest

Quick Example

using Conjecture.Xunit;

public class SortTests
{
    [Property]
    public bool Sorting_is_idempotent(List<int> items)
    {
        List<int> sorted = items.OrderBy(x => x).ToList();
        List<int> sortedTwice = sorted.OrderBy(x => x).ToList();
        return sorted.SequenceEqual(sortedTwice);
    }
}

Run with dotnet test. Conjecture generates random lists, runs the property 100 times, and if it fails, shrinks the input to the minimal failing case.

Features

• Automatic test generation — generates random inputs from type-aware strategies
• Intelligent shrinking — finds the smallest failing input via byte-stream minimization
• LINQ composition — build complex strategies with Select, Where, SelectMany
• All major test frameworks — xUnit v2, xUnit v3, NUnit, MSTest, Microsoft Testing Platform
• F# support — idiomatic Gen<'a> wrapper plus an Expecto integration
• Source generators — derive strategies for your types with [Arbitrary]
• Roslyn analyzers — catch common mistakes at compile time
• Stateful testing — model systems as state machines and explore command sequences
• Targeted testing — steer generation toward extremes with Target.Maximize / Target.Minimize
• Recursive strategies — generate bounded-depth trees and self-referential types
• Example database — persist failing inputs for automatic regression prevention
• Reproduction export — write a runnable .cs repro of any failing input via ExportReproductionOnFailure
• Structured logging — structured events for generation, shrinking, and targeting phases
• HTTP / gRPC / messaging / EF Core / ASP.NET Core / Aspire integrations — model real distributed systems as IInteractionTargets and dispatch composed interactions through one state machine
• MCP server + dotnet tool — generate strategies and standalone test data from the CLI or AI assistants

Packages

Documentation

Full documentation is at ommundsen.dev/Conjecture:

• Quick Start — write your first property test in 5 minutes
• Tutorials — learn property-based testing step by step
• API Reference — auto-generated from source
• Porting Guide — coming from Python Hypothesis?
• Changelog

Credit

Conjecture is a .NET implementation of Hypothesis's property-based testing engine. The shrinking algorithm, choice-sequence IR, targeting, and example database all derive from the work of David R. MacIver, Zac Hatfield-Dodds, and the broader Hypothesis contributors. The C# API surface, source generators, and analyzer integrations are original to this project.

License

Source code: MPL-2.0 | NuGet packages: MIT