RoslynLens

Quick Start

Get RoslynLens running with Claude Code in under 2 minutes.

1. Install

dotnet tool install --global RoslynLens

2. Register with Claude Code

claude mcp add --scope user --transport stdio roslyn-lens -- roslyn-lens

3. Try it

Open a .NET project directory and start Claude Code. The MCP server auto-discovers your .sln or .slnx file (BFS up to 3 levels).

Find a type

"Find the WorkspaceManager class"

Claude Code calls find_symbol and gets back ~50 tokens instead of reading the entire file (~1500 tokens).

Inspect the public API

"What's the public API of IAntiPatternDetector?"

Claude Code calls get_public_api and receives a concise list of members.

Check complexity

"Which methods have the highest cyclomatic complexity in this project?"

Claude Code calls get_complexity_metrics with scope=project and gets a sorted list of hotspots.

Detect anti-patterns

"Run the anti-pattern detectors on Program.cs"

Claude Code calls detect_antipatterns scoped to the file and gets violations with suggestions.

Get a full method analysis

"Analyze the ExecuteAsync method in FindSymbolTool"

Claude Code calls analyze_method and receives signature, callers, dependency graph, and complexity metrics — all in a single round-trip.

How it works

Claude Code ──stdin/stdout──▶ RoslynLens MCP Server ──▶ Roslyn SemanticModel
     │                              │
     │  JSON-RPC request            │  Compilation + symbol resolution
     │  (tool name + params)        │  LRU cache (configurable)
     │                              │
     ◀──── JSON response ◀──────────┘
           (30-150 tokens)

Instead of Claude Code reading .cs files (500-2000+ tokens each), it queries RoslynLens and receives focused, semantic results. This dramatically reduces token consumption on large .NET solutions.

Next steps