Advanced Analysis Tools

These tools provide deeper code analysis capabilities beyond navigation and inspection: complexity metrics, data flow, control flow, duplicate detection, and external source resolution.

get_complexity_metrics

Compute cyclomatic complexity, cognitive complexity, maximum nesting depth, and logical lines of code for methods.

Parameters

Name	Type	Required	Description
name	string	yes	Method, type, or project name
scope	string	no	method, type, or project (default method)
className	string	no	Containing class for method disambiguation
threshold	int	no	Minimum cyclomatic complexity to include (default 0)
maxResults	int	no	Max results (default 50)

Metrics explained

Metric	What it measures	Threshold guidance
Cyclomatic	Number of linearly independent paths. Counts if, while, for, foreach, case, &&, ||, ??, catch, ternary. Base = 1.	< 10 good, 10-20 moderate, > 20 high risk
Cognitive	Human-readability cost (SonarSource model). Nesting structures add 1 + nesting level as penalty. else if does not increase nesting.	< 15 good, 15-30 moderate, > 30 refactor
Max nesting	Deepest nesting level of control flow blocks.	< 3 good, 4 yellow, 5+ refactor
Logical LOC	Non-blank, non-comment lines.	< 30 good, 30-60 moderate, > 60 split

Example: scan a project for hotspots

name: "MyProject"
scope: "project"
threshold: 10

Returns methods sorted by cyclomatic complexity (descending), filtered to those above the threshold.

{
  "Methods": [
    {
      "Method": "ProcessOrder",
      "ContainingType": "OrderService",
      "File": "OrderService.cs",
      "Line": 45,
      "Metrics": { "Cyclomatic": 18, "Cognitive": 24, "MaxNesting": 4, "LogicalLoc": 52 }
    },
    {
      "Method": "ValidateInput",
      "ContainingType": "InputValidator",
      "File": "InputValidator.cs",
      "Line": 12,
      "Metrics": { "Cyclomatic": 12, "Cognitive": 15, "MaxNesting": 3, "LogicalLoc": 28 }
    }
  ],
  "Total": 2
}

---

analyze_data_flow

Analyze variable data flow within a method using Roslyn's SemanticModel.AnalyzeDataFlow().

Parameters

Name	Type	Required	Description
methodName	string	yes	Name of the method
className	string	no	Containing class to disambiguate

Response fields

Field	Description
VariablesDeclared	Variables declared inside the method body
DataFlowsIn	Variables whose values flow into the region from outside
DataFlowsOut	Variables whose values flow out of the region
ReadInside	Variables read inside the method
WrittenInside	Variables written inside the method
AlwaysAssigned	Variables always assigned before the method exits
Captured	Variables captured by lambdas or local functions

When to use

• Understanding side effects — which variables escape the method?
• Refactoring — can a block be safely extracted?
• Lambda analysis — which variables are captured (closures)?

---

analyze_control_flow

Analyze control flow within a method using Roslyn's SemanticModel.AnalyzeControlFlow().

Parameters

Name	Type	Required	Description
methodName	string	yes	Name of the method
className	string	no	Containing class to disambiguate

Response fields

Field	Description
StartReachable	Whether the first statement is reachable
EndReachable	Whether the end of the method is reachable (no guaranteed return/throw)
ReturnStatements	Number of return statements
ExitPoints	Total exit points (returns, throws, breaks)

When to use

• Dead code detection — is the end of a method reachable?
• Complexity assessment — how many exit points?
• Refactoring — ensuring all paths return

---

detect_duplicates

Detect structurally similar code blocks using AST fingerprinting.

How it works

1. Extract all method bodies as statement sequences
2. Normalize each: replace identifiers with ID, literals with LIT (creates a structural fingerprint)
3. Hash the normalized form (SHA-256)
4. Group methods with identical hashes (exact structural duplicates)

Two methods that differ only in variable names or literal values will produce the same fingerprint.

Parameters

Name	Type	Required	Description
projectFilter	string	no	Scope to a specific project
minStatements	int	no	Minimum statements to consider (default 5)
maxResults	int	no	Maximum duplicate groups (default 20)

Example response

{
  "Groups": [
    {
      "Occurrences": [
        { "Method": "ValidateEmail", "ContainingType": "UserService", "File": "UserService.cs", "Line": 34 },
        { "Method": "ValidatePhone", "ContainingType": "ContactService", "File": "ContactService.cs", "Line": 22 }
      ],
      "Similarity": 1.0,
      "StatementCount": 8
    }
  ],
  "TotalGroups": 1,
  "TotalDuplicates": 2
}

When to use

• Before a refactoring sprint — find extraction candidates
• Code review — spot copy-paste patterns
• Technical debt assessment — quantify duplication

---

resolve_external_source

Resolve source code for external (NuGet/framework) symbols via SourceLink or decompilation.

Parameters

Name	Type	Required	Description
symbolName	string	yes	External symbol name
assemblyName	string	no	Assembly name to narrow search

Resolution hierarchy

1. SourceLink — if the symbol has source in the solution or SourceLink metadata in the PDB
2. Decompilation — using ICSharpCode.Decompiler as a fallback
3. None — if neither method succeeds

Example response

{
  "Symbol": "Microsoft.Extensions.Logging.ILogger",
  "Assembly": "Microsoft.Extensions.Logging.Abstractions",
  "SourceLinkUrl": null,
  "DecompiledSource": "using System;\n\nnamespace Microsoft.Extensions.Logging\n{\n    public interface ILogger\n    {\n        ...\n    }\n}",
  "ResolutionMethod": "decompilation"
}

Decompiled source is truncated to 60 lines for token efficiency.

When to use

• Understanding a third-party API's implementation details
• Debugging framework behavior without leaving the MCP server
• Verifying NuGet package behavior