KnockOff 0.54.0
dotnet add package KnockOff --version 0.54.0
NuGet\Install-Package KnockOff -Version 0.54.0
<PackageReference Include="KnockOff" Version="0.54.0" />
<PackageVersion Include="KnockOff" Version="0.54.0" />
<PackageReference Include="KnockOff" />
paket add KnockOff --version 0.54.0
#r "nuget: KnockOff, 0.54.0"
#:package KnockOff@0.54.0
#addin nuget:?package=KnockOff&version=0.54.0
#tool nuget:?package=KnockOff&version=0.54.0
KnockOff
A .NET mocking library that lets you define reusable stub classes — with full mocking capabilities built in.
Define your test double once. Reuse it across your test project. Customize it per-test with Return, Call, Verify, and When chains. No more copying mock setups between tests or maintaining shared factory methods full of Arg.Any<>().
Powered by Roslyn source generation for tighter type safety — more issues surface as compile errors instead of runtime surprises.
Claude Code was used to write this library. Skip to more AI discussion.
We've confirmed KnockOff matches Moq and NSubstitute performance across 5,000 unit tests — build and execution — even with source generation overhead.
KnockOff Stub
There are 9 patterns total, including a standard fluent mocking approach with inline stubs. But reusable stub classes are where KnockOff stands apart:
[KnockOff]
public partial class MyRepoStub(List<User> Users) : IMyRepo
{
protected override User? GetUser_(int id)
{
return Users.Single(u => u.Id == id);
}
protected override void Update_(User user)
{
Assert.Contains(user, Users);
}
}
[KnockOff]+partial class— KnockOff generates a base class that implements every member ofIMyRepo. Your stub is a real class — define it once, reuse it across your entire test project. Pass it around, register it in DI, share it between test fixtures.- Constructor parameters —
List<User> Usersis a primary constructor. Test data flows in naturally, just like any other C# class. - Overrides are optional —
GetUser_andUpdate_override the generated defaults. Only override what you need — everything else still works with Return/Call, Return(value), or When chains. - Tighter type safety — Every Return, Call, and When call is complete in a single step — no forgotten
.Returns()that silently breaks at runtime. No manual<T1, T2>type parameters that can drift. Details →
This stub is also a full mock. It has Verify, Strict mode, Async, and Source Delegation — all on the same reusable class.
Why I Wrote KnockOff
I often wanted to reuse my mocks. Especially in my integration test library where I may even register my mocks. I found myself either copying my mock definitions code or creating shared methods like this:
NSubstitute:
public static IMyRepo NSubstituteMock(List<User> users)
{
var myRepoMock = Substitute.For<IMyRepo>();
// Setup: configure GetUser to look up from the list based on id
myRepoMock.GetUser(Arg.Any<int>())
.Returns(callInfo => users.SingleOrDefault(u => u.Id == callInfo.Arg<int>()));
// Setup: configure Update to assert user exists in list
myRepoMock.When(x => x.Update(Arg.Any<User>()))
.Do(callInfo => Assert.Contains(callInfo.Arg<User>(), users));
return myRepoMock;
}
Here's another example from PowerToys.
But I find that hard to read and unintuitive. Also, my shared methods accumulated extra parameters for variations across different tests.
So I Created KnockOff
You can create a stub to implement interfaces or non-sealed classes with virtual methods. Yet, you can still customize the stub per test. All while having the features you would expect with a full mocking library.
With the stub above, your tests are:
var myRepoKO = new MyRepoStub([new User { Id = 1 }, new User { Id = 2 }]);
var userDomainModel = new UserDomainModel(myRepoKO);
Assert.True(userDomainModel.Fetch(1));
// I have Verify on my Stub!
myRepoKO.GetUser.Verify(Called.Once);
Need different behavior for a specific test? Override with Return/Call:
var user1 = new User { Id = 1 }; // Ignored do to per-test configuration
var myRepoKO = new MyRepoStub([user1]);
var userDomainModel = new UserDomainModel(myRepoKO);
var user2 = new User { Id = 2 };
// When and Return overrides the stub methods
myRepoKO.GetUser.When(2).Return(user2).Verifiable();
myRepoKO.Update.Call(u => Assert.Same(u, user2)).Verifiable();
userDomainModel.Fetch(2);
userDomainModel.Update();
myRepoKO.Verify();
Now I have my stubs and mocks in one!
What Sets KnockOff Apart
- Reusable stub classes — Define once, customize per-test. Your stub is a real class — pass it through constructors, register it in DI.
- Source delegation — Delegate to a real implementation, override only specific methods. No equivalent in Moq or NSubstitute.
- Protected methods — Same
Return/Call/VerifyAPI, fully typed. No string-based names, no manual subclasses. - Ref/out parameters — Natural lambda syntax with
ref/outkeywords. No special matchers or index-based access. - Multiple interfaces — Unified interceptors on one stub. No
.As<T>()references or casting. - Tighter type safety — Each Return/Call/When call is complete in one step — no forgotten
.Returns()that silently breaks at runtime. - Parameter matching —
Call((int a, int b) => a > 0 ? 100 : 0)— standard C# conditionals instead ofArg.Is<>orIt.Is<>per parameter. - Built-in argument capture —
LastArg,LastArgs,LastSetValue,LastSetEntry— no manualArg.Do<>orCallback<>setup. - Event verification —
VerifyAdd()/VerifyRemove()/HasSubscribers— not available in Moq or NSubstitute. - Explicit Get/Set verification —
VerifyGet(Called)/VerifySet(Called)for properties and indexers. - Stubbing concrete classes — Override virtual methods on non-sealed classes with the same API.
Performance
Source generation adds code to your build, so we wanted to confirm KnockOff keeps up with Moq and NSubstitute. We ran 5,000 equivalent unit tests across all three frameworks.
Test Execution (5,000 tests each)
| Project | Duration |
|---|---|
| KnockOff Inline Stubs | ~600ms |
| KnockOff Standalone Stubs | ~610ms |
| NSubstitute | ~870ms |
| Moq | ~1,000ms |
Clean Build Time
| Project | Duration |
|---|---|
| NSubstitute | 2.83s |
| KnockOff Inline Stubs | 3.35s |
| KnockOff Standalone Stubs | 3.61s |
| Moq | 4.57s |
KnockOff's build times land between NSubstitute and Moq despite generating all stubs at compile time. Test execution is comparable across all three frameworks. Full methodology and test design in the 5000 Unit Tests repository.
Quick Start
Install
dotnet add package KnockOff
Create a Stub
public interface IQuickStartRepo
{
User? GetUser(int id);
}
[KnockOff]
public partial class QuickStartRepoStub : IQuickStartRepo { }
public class QuickStartCreateStubTests
{
[Fact]
public void CreateStub_IsReady()
{
var stub = new QuickStartRepoStub();
IQuickStartRepo repository = stub;
Assert.NotNull(repository);
}
}
Configure and Verify
[Fact]
public void ConfigureStub_WithReturn()
{
var stub = new QuickStartRepoStub();
stub.GetUser.Call((id) => new User { Id = id, Name = "Test User" });
IQuickStartRepo repository = stub;
var user = repository.GetUser(42);
Assert.NotNull(user);
Assert.Equal(42, user.Id);
Assert.Equal("Test User", user.Name);
}
[Fact]
public void VerifyCalls_WithVerifiable()
{
var stub = new QuickStartRepoStub();
stub.GetUser.Call((id) => new User { Id = id, Name = "Test" }).Verifiable();
IQuickStartRepo repository = stub;
var user = repository.GetUser(42);
// Verify() checks all members marked with .Verifiable()
stub.Verify();
}
The Difference
Moq:
mock.Setup(x => x.GetUser(It.Is<int>(id => id > 0)))
.Returns<int>(id => new User { Id = id });
NSubstitute:
var repo = Substitute.For<IUserRepo>();
repo.GetUser(Arg.Is<int>(id => id > 0)).Returns(x => new User { Id = x.Arg<int>() });
KnockOff:
var stub = new CompareUserRepoStub();
stub.GetUser.Call((id) => id > 0 ? new User { Id = id } : null);
No It.Is<>(). No Arg.Is<>(). No x.Arg<int>(). The parameter is just id.
For side-by-side comparison tables (methods, properties, events, delegates, indexers), see the complete comparison guide.
Argument Matching
Moq:
// Moq - It.Is<T> per parameter
mock.Setup(x => x.Add(It.Is<int>(a => a > 0), It.IsAny<int>())).Returns(100);
NSubstitute:
// NSubstitute - Arg.Is<T> per parameter (permanent matchers)
calc.Add(Arg.Is<int>(a => a > 0), Arg.Any<int>()).Returns(100);
KnockOff:
// KnockOff - Returns with conditional (permanent, matches all calls)
stub.Add.Call((int a, int b) => a > 0 ? 100 : 0);
// KnockOff - When() for sequential matching (first match returns 100, then falls through)
stub.Add.When((int a, int b) => a > 0).Return(100).ThenCall((int a, int b) => a + b);
Multiple specific values:
Moq:
mock.Setup(x => x.Add(1, 2)).Returns(100);
mock.Setup(x => x.Add(3, 4)).Returns(200);
// Multiple specific values
calc.Add(1, 2).Returns(100);
calc.Add(3, 4).Returns(200);
stub.Add.When(1, 2).Return(100);
stub.Add.When(3, 4).Return(200);
Note: Moq and NSubstitute matchers are permanent -- they match all qualifying calls. KnockOff's When() is sequential -- matchers are consumed in order. Use Call(callback) with conditionals for permanent matching behavior.
Argument Capture
Moq:
// Moq - requires Callback setup
int capturedA = 0, capturedB = 0;
mock.Setup(x => x.Add(It.IsAny<int>(), It.IsAny<int>()))
.Callback<int, int>((a, b) => { capturedA = a; capturedB = b; });
mock.Object.Add(1, 2);
NSubstitute:
// NSubstitute - requires Arg.Do in setup
int capturedA = 0, capturedB = 0;
calc.Add(Arg.Do<int>(x => capturedA = x), Arg.Do<int>(x => capturedB = x));
calc.Add(1, 2);
KnockOff:
// KnockOff - built-in, no pre-setup
var tracking = stub.Add.Call((int a, int b) => a + b);
ICalculator calc = stub;
calc.Add(1, 2);
var (a, b) = tracking.LastArgs; // Named tuple: a = 1, b = 2
For full comparisons of properties, events, delegates, and indexers, see the complete comparison guide.
Method Overload Resolution
The Problem: When an interface has overloaded methods with the same parameter count but different types:
public interface IFormatter
{
string Format(string input, bool uppercase);
string Format(string input, int maxLength);
}
Any-Value Matching
Moq:
// It.IsAny<T>() required - compiler needs the types to resolve overload
mock.Setup(x => x.Format(It.IsAny<string>(), It.IsAny<bool>())).Returns("bool overload");
mock.Setup(x => x.Format(It.IsAny<string>(), It.IsAny<int>())).Returns("int overload");
NSubstitute:
// Arg.Any<T>() required - compiler needs the types to resolve overload
formatter.Format(Arg.Any<string>(), Arg.Any<bool>()).Returns("bool overload");
formatter.Format(Arg.Any<string>(), Arg.Any<int>()).Returns("int overload");
KnockOff:
// Explicit parameter types resolve the overload - standard C# syntax
stub.Format.Call((string input, bool uppercase) => "bool overload");
stub.Format.Call((string input, int maxLength) => "int overload");
Specific-Value Matching
NSubstitute:
// Specific value matching - literals work when all args are specific
formatter.Format("test", true).Returns("UPPERCASE");
formatter.Format("test", 10).Returns("truncated");
KnockOff:
// Specific value matching - parameter types resolve the overload
stub.Format.When("test", true).Return("UPPERCASE");
stub.Format.When("test", 10).Return("truncated");
Argument Access
Moq:
// To use argument values, extract via Returns<T1, T2>:
mock.Setup(x => x.Format(It.IsAny<string>(), It.IsAny<bool>()))
.Returns<string, bool>((input, uppercase) => uppercase ? input.ToUpper() : input);
NSubstitute:
// To use argument values, extract from CallInfo:
formatter.Format(Arg.Any<string>(), Arg.Any<bool>())
.Returns(x => x.ArgAt<bool>(1) ? x.ArgAt<string>(0).ToUpper() : x.ArgAt<string>(0));
KnockOff:
// Arguments are directly available with names and types:
stub.Format.Call((string input, bool uppercase) => uppercase ? input.ToUpper() : input);
The Difference:
- Moq:
It.IsAny<bool>()+.Returns<string, bool>((input, uppercase) => ...)to match any value and access arguments - NSubstitute:
Arg.Any<bool>()+x.ArgAt<bool>(1)to match any value and access arguments - KnockOff:
(string input, bool uppercase)- standard C# lambda with named, typed parameters
Three Stub Patterns
KnockOff supports 9 patterns total. Here are the three most common:
Standalone - Reusable across your project:
[KnockOff]
public partial class ReadmeStandaloneStub : IUserRepo { }
Inline Interface - Test-local stubs:
[Fact]
public void InlineInterface_Pattern()
{
var stub = new Stubs.IUserRepo();
stub.GetUser.Call((id) => new User { Id = id });
IUserRepo repo = stub;
Assert.NotNull(repo.GetUser(1));
}
Inline Class - Stub virtual members:
[Fact]
public void InlineClass_Pattern()
{
var stub = new Stubs.MyService();
stub.GetUser.Call((id) => new User { Id = id });
MyService service = stub.Object;
Assert.NotNull(service.GetUser(1));
}
Roslyn Source Generation
KnockOff uses Roslyn source generation, which means:
- No more
Arg.Any<>(). No moreIt.IsAny<>(). Just write C# - If the method signature changes you get a compile error
- There's a small performance gain but honestly it's negligible
Source generation opens doors beyond traditional mocking — I've already added 9 patterns and features like Source Delegation, with more ideas to come.
What other ideas do you have? Open a discussion.
AI
This is an idea I've had for years but never took the time to implement. With my ideas and guidance, Claude Code has written the entirety of this library — the Roslyn source generator, the runtime library, the tests, and the documentation.
Source generation turned out to be a great fit for AI code generation. The work is highly patterned: analyze an interface, generate code for each member, handle edge cases across 9 patterns and 4 member types. That's exactly the kind of systematic, repetitive-but-varied work where AI excels. I designed the API and patterns; Claude Code implemented them across every combination.
Claude Code Skill
KnockOff includes a Claude Code skill that teaches Claude how to use the library. Copy the skills/knockoff/ directory into your project and Claude Code will know how to create stubs, configure behavior, write tests with KnockOff, and migrate from Moq — without you explaining the API.
The skill includes slash commands:
/knockoff:create-stub— Create a new stub class with the pattern of your choice/knockoff:migrate-from-moq— Convert existing Moq tests to KnockOff/knockoff:troubleshoot— Diagnose and fix common KnockOff issues
Documentation
- Getting Started - Installation and first stub
- Stub Patterns - Standalone, inline interface, inline class
- Interceptor API - Complete
Return,Call,Get,Setreference - Source Delegation - Delegate to real implementations
- Full Comparison Guide - Properties, events, delegates, indexers vs Moq and NSubstitute
- Migration from Moq - Step-by-step migration guide
- Migration from NSubstitute - Comparison and migration guide
- Migration to v0.52 - API rename guide (Return/Call/When)
License
MIT License. See LICENSE for details.
Contributing
Contributions welcome! See CONTRIBUTING.md for guidelines.
- Issues: GitHub Issues
- Pull Requests: Bug fixes, features, documentation
- Discussions: GitHub Discussions
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net8.0 is compatible. net8.0-android was computed. net8.0-browser was computed. net8.0-ios was computed. net8.0-maccatalyst was computed. net8.0-macos was computed. net8.0-tvos was computed. net8.0-windows was computed. net9.0 is compatible. net9.0-android was computed. net9.0-browser was computed. net9.0-ios was computed. net9.0-maccatalyst was computed. net9.0-macos was computed. net9.0-tvos was computed. net9.0-windows was computed. net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- No dependencies.
-
net8.0
- No dependencies.
-
net9.0
- No dependencies.
NuGet packages
This package is not used by any NuGet packages.
GitHub repositories
This package is not used by any popular GitHub repositories.
| Version | Downloads | Last Updated |
|---|---|---|
| 0.54.0 | 81 | 2/20/2026 |
| 0.52.0 | 91 | 2/18/2026 |
| 0.51.2 | 84 | 2/16/2026 |
| 0.51.1 | 130 | 2/16/2026 |
| 0.51.0 | 88 | 2/16/2026 |
| 0.49.0 | 91 | 2/14/2026 |
| 0.48.0 | 87 | 2/10/2026 |
| 0.47.0 | 85 | 2/10/2026 |
| 0.46.0 | 91 | 2/9/2026 |
| 0.45.0 | 83 | 2/9/2026 |
| 0.44.0 | 88 | 2/8/2026 |
| 0.43.0 | 82 | 2/8/2026 |
| 0.42.0 | 91 | 2/8/2026 |
| 0.40.0 | 90 | 2/8/2026 |
| 0.37.0 | 97 | 2/7/2026 |
| 0.36.0 | 83 | 2/6/2026 |
Fix: stub override on partial overloads no longer splits interceptor groups. All overloads share a single interceptor with per-signature stub override fallback.