DotnetCodingContextMCP 0.3.1

DotnetCodingContextMCP

An MCP server that gives AI agents the coding context they need to write compilable C# code — type signatures, constructors, mock patterns, and test conventions in a single call.

The Problem

When AI agents write C# code, they hallucinate type names, property names, constructor signatures, and assertion patterns:

This causes build-fail-fix-build spirals that waste time and tokens.

The Solution: 4 Tools

analyze_solution_for_testing (new in v0.2.0)

Discover what's testable in a solution — call this FIRST.

Input:  { "path": "/src/MyApp.sln" }

Output:
{
  "sourceProjects": ["MyApp.Core", "MyApp.API"],
  "testProjects": ["MyApp.Tests"],
  "projects": [
    {
      "projectName": "MyApp.Core",
      "totalPublicTypes": 42,
      "testableTypes": [
        {
          "name": "UserService",
          "kind": "class",
          "publicMethodCount": 8,
          "constructorDependencies": ["IUserRepository", "ILogger<UserService>"],
          "complexity": 12,
          "instantiationHint": "new UserService(Mock.Of<IUserRepository>(), Mock.Of<ILogger<UserService>>())"
        }
      ]
    }
  ],
  "recommendedTestSetup": "Most-used dependencies to mock: IUserRepository, ILogger. Create mock instances first."
}

get_coding_context

One call → everything needed to write code against specific types.

Input:  { "typeNames": "UserService, OrderRepository",
          "path": "/src/MyApp.sln" }

Output:
{
  "types": [
    {
      "name": "UserService",
      "using": "using MyApp.Services;",
      "kind": "class",
      "constructors": ["new UserService(IUserRepository repo, ILogger<UserService> logger)"],
      "properties": [
        { "name": "IsInitialized", "type": "bool", "settable": false }
      ],
      "instantiation": "new UserService(Mock.Of<IUserRepository>(), Mock.Of<ILogger<UserService>>())"
    }
  ],
  "relatedTypes": [...]
}

get_interface_contract

Mock-ready interface extraction with Moq setup patterns.

Input:  { "interfaceName": "IUserRepository", "path": "/src/MyApp.sln" }

Output:
{
  "members": [
    {
      "name": "GetByIdAsync",
      "signature": "Task<User?> GetByIdAsync(int id)",
      "mockSetup": "mock.Setup(x => x.GetByIdAsync(It.IsAny<int>())).ReturnsAsync(default(User?)!)"
    }
  ],
  "implementations": ["SqlUserRepository", "InMemoryUserRepository"],
  "parameterTypes": [{ "name": "User", "kind": "class", "instantiation": "..." }]
}

get_project_test_context

Analyze existing tests to extract conventions.

Input:  { "testProjectPath": "/test/MyApp.Tests/MyApp.Tests.csproj" }

Output:
{
  "framework": "xunit",
  "assertionLibrary": "Shouldly",
  "mockingLibrary": "Moq",
  "assertionPatterns": {
    "equality": "result.ShouldBe(expected)",
    "true": "result.ShouldBeTrue()",
    "null": "result.ShouldBeNull()"
  },
  "namingConvention": "MethodName_Scenario_ExpectedBehavior",
  "helperMethods": [
    { "name": "CreateTestUser", "signature": "User CreateTestUser(string name)" }
  ]
}

Quick Start

Install

dotnet tool install --global DotnetCodingContextMCP

VS Code / GitHub Copilot

Add to mcp.json:

{
  "servers": {
    "coding-context": {
      "type": "stdio",
      "command": "coding-context-mcp"
    }
  }
}

Pair with DotnetTemplateMCP

For complete .NET AI tooling — templates for project creation, coding context for code authoring:

{
  "servers": {
    "dotnet-templates": {
      "type": "stdio",
      "command": "template-engine-mcp"
    },
    "coding-context": {
      "type": "stdio",
      "command": "coding-context-mcp"
    }
  }
}

MSBench / GitHub Copilot CLI with Custom Agent

For benchmark runs, combine the MCP server with a custom agent prompt to instruct the agent to use the tools:

# In your runner.sh, before sourcing entry.sh:

# 1. Configure MCP server
export GITHUB_COPILOT_MCP_JSON='{"mcpServers":{"coding-context":{"type":"local","tools":["*"],"command":"coding-context-mcp","args":[]}}}'

# 2. Create custom agent with MCP usage instructions
mkdir -p ~/.copilot/agents
cat > ~/.copilot/agents/dotnet-test-writer.md << 'EOF'
---
name: dotnet-test-writer
description: Expert .NET test writer that uses coding context MCP tools.
tools: ["*"]
---

You have access to coding-context MCP tools. Follow this workflow:

1. FIRST call `coding-context-analyze_solution_for_testing` with the solution path
2. Before writing code that uses specific types, call `coding-context-get_coding_context`
3. Before mocking interfaces, call `coding-context-get_interface_contract`
4. Before creating test files, call `coding-context-get_project_test_context`
EOF

export COPILOT_CUSTOM_AGENT=dotnet-test-writer

See examples/ for complete msbench runner scripts.

Why Not SharpLensMcp / Other Roslyn MCPs?

Existing Roslyn MCP servers expose many granular tools designed for code exploration (navigate, inspect, analyze). This MCP exposes 4 compound tools designed for code authoring — giving agents exactly what they need to write compilable code:

Requirements

• .NET 10 SDK
• MSBuild (included with .NET SDK)

License

MIT