USAGE.md

DLL Decompiler MCP Server - Usage Examples

Quick Start

1. Start the Server

cd c:\development\mcpeek\McPeek
dotnet run

The server will start and listen for MCP requests via stdio.

2. Configure in VS Code/Cline

Add to your MCP settings file (usually in VS Code settings or Cline configuration):

{
  "mcpServers": {
    "mcpeek": {
      "command": "dotnet",
      "args": [
        "run",
        "--project",
        "c:\\development\\mcpeek\\DllDecompilerMcp\\DllDecompilerMcp.csproj"
      ]
    }
  }
}

Example Workflows

Example 1: Decompile a Project's Dependencies

AI: Use the DecompileFolder tool to decompile all DLLs in C:\MyProject\bin\Debug\net8.0

Expected Response:

{
  "Success": true,
  "Message": "Successfully decompiled 15 assemblies",
  "AssemblyNames": [
    "MyProject",
    "Microsoft.Extensions.DependencyInjection",
    ...
  ],
  "TotalFiles": 450
}

Example 2: Find Implementation of an Interface

AI: Search for "IServiceProvider" in the decompiled code

Expected Response:

{
  "Success": true,
  "Query": "IServiceProvider",
  "TotalMatches": 23,
  "Results": [
    {
      "AssemblyName": "Microsoft.Extensions.DependencyInjection",
      "FilePath": "ServiceProvider.cs",
      "LineNumber": 15,
      "LineContent": "public class ServiceProvider : IServiceProvider",
      "Context": "...(surrounding code)..."
    },
    ...
  ]
}

Example 3: Analyze a Specific Class

AI: First, decompile C:\MyProject\bin\Debug\MyLibrary.dll
Then search for "MyImportantClass"

This workflow:

1. Decompiles the specific DLL
2. Searches for the class
3. AI can then analyze the implementation details

Example 4: Understanding Dependencies

AI: List all loaded assemblies

Expected Response:

{
  "Success": true,
  "TotalAssemblies": 15,
  "Assemblies": [
    {
      "Name": "MyProject",
      "Path": "C:\\MyProject\\bin\\Debug\\MyProject.dll",
      "FileCount": 30,
      "DecompiledAt": "2025-10-24T12:00:00Z"
    },
    ...
  ]
}

Example 5: Cache Management

AI: Get cache statistics

Expected Response:

{
  "Success": true,
  "TotalCachedAssemblies": 15,
  "CacheDirectory": "C:\\Users\\User\\AppData\\Local\\Temp\\DllDecompilerMcp\\cache",
  "CacheSizeMB": 12.5
}

To clear the cache:

AI: Clear the decompilation cache

Real-World Scenarios

Scenario 1: Understanding Third-Party Libraries

Objective: Understand how a third-party NuGet package implements a specific feature.

Steps:

1. Locate the DLL in your NuGet packages folder (usually %USERPROFILE%\.nuget\packages)
2. Decompile: DecompileDll C:\Users\You\.nuget\packages\SomeLibrary\1.0.0\lib\net8.0\SomeLibrary.dll
3. Search for the feature: SearchCode "FeatureName"
4. Ask AI to explain the implementation

Scenario 2: Debugging Production Issues

Objective: Analyze decompiled code from a production DLL to understand a bug.

Steps:

1. Copy production DLLs to a local folder
2. Decompile the folder: DecompileFolder C:\ProductionDlls
3. Search for the problematic method: SearchCode "ProblematicMethod"
4. Ask AI to identify potential issues in the code

Scenario 3: API Documentation Generation

Objective: Generate documentation for an undocumented internal library.

Steps:

1. Decompile the library: DecompileDll C:\InternalLibs\MyLib.dll
2. List all types: Ask AI to summarize all public classes
3. For each class, ask AI to generate documentation based on the decompiled code

Scenario 4: Migration Analysis

Objective: Understand what changes are needed to migrate from one framework version to another.

Steps:

1. Decompile old framework DLLs: DecompileFolder C:\OldFramework\bin
2. Decompile new framework DLLs: DecompileFolder C:\NewFramework\bin
3. Search for specific APIs: SearchCode "ObsoleteAPI"
4. Ask AI to identify migration paths

Tips

1. Cache is Your Friend: The cache uses SHA256 hashes, so if you decompile the same DLL multiple times, it's instant after the first time.

2. Search Smartly: Use specific search terms to reduce the number of results. The default limit is 50 results per search.

3. Case Sensitivity: By default, searches are case-insensitive. Use the caseSensitive parameter if you need exact matches.

4. Large Projects: For large projects with many DLLs, decompile one folder at a time to keep things manageable.

5. XML Documentation: The decompiler includes XML documentation comments when available, making the AI's understanding more accurate.

Troubleshooting

Issue: "Failed to decompile DLL"

Possible causes:

• The file is not a .NET assembly
• The assembly is obfuscated
• File permissions issue

Solution:

• Verify the file is a valid .NET DLL using dotnet --info or ILSpy
• Check file permissions

Issue: "Cache growing too large"

Solution:

AI: Clear the decompilation cache

Then selectively decompile only what you need.

Issue: "Search returns too many results"

Solution:

• Use more specific search terms
• Add the namespace or class name
• Use the maxResults parameter to limit results
• Use caseSensitive search for exact matches

Performance Notes

• First decompilation: Can take a few seconds per DLL depending on size
• Cached decompilation: Instant (< 100ms)
• Search: Very fast across all loaded assemblies
• Cache location: %TEMP%\McPeek\cache

Security Considerations

• The server only decompiles local DLLs that you explicitly specify
• No network access required
• Cache is stored in your local temp folder
• All operations are read-only (doesn't modify source DLLs)

Integration with AI Workflows

The MCP server is designed to work seamlessly with AI assistants like:

• GitHub Copilot (VS Code Agent Mode)
• Claude with MCP support
• Custom AI agents built with Semantic Kernel

The AI can:

• Understand complex library implementations
• Explain how specific features work
• Identify patterns and best practices
• Suggest improvements based on decompiled code
• Help with debugging by analyzing actual compiled code