RoslynMcp

A Model Context Protocol (MCP) server that brings Roslyn code intelligence to AI agents.

Get It on NuGet

This project uses Roslynator, licensed under Apache 2.0.

Installation

dotnet tool install -g RoslynMcp

Update

dotnet tool update -g RoslynMcp

MCP config (OpenCode)

  "mcp": {
    "code inspection": {
      "type": "local",
      "command": [
        "roslynmcp"
      ]
    }
  }

What It Is

RoslynMcp is a .NET application that exposes the power of Roslyn (the .NET compiler platform) through the MCP protocol. It acts as a bridge between AI assistants and your C# codebase, enabling deep code understanding and analysis.

Why It Exists

Traditional AI code assistants often rely on simplistic pattern matching (grep/glob) which misses semantic context. RoslynMcp solves this by providing:

• Semantic understanding — It knows what your code means, not just what it says
• Symbol resolution — Understands types, methods, properties across your entire solution
• Call graph tracing — See how code flows through your system
• Code smell detection — Identifies potential issues using Roslynator analyzers

What You Can Use It For

Inspection Tool	Description
Load Solution	Loads a .sln/.slnx file and prepare the workspace
Understand Codebase	Quick orientation with complexity hotspots
List Dependencies	Understand project relationships
List Types	Discover all classes, interfaces, enums in a project
List Members	Explore methods, properties, fields of any type
Resolve Symbol	Get canonical IDs for a single code symbol
Resolve Symbols	Resolve multiple symbols in one round-trip
Explain Symbol	Understand what a symbol does and where it's used
Trace Call Flow	See upstream callers or downstream callees
Find Callers	Return only immediate direct upstream callers
Find Callees	Return only immediate direct downstream callees
Find Usages	Locate all references to a type/member
Find Implementations	Locate all implementations of a interface or abstract class/method
Get Type Hierarchy	Explore type inheritance and derived types
Find Code Smells	Detect potential issues in a file

Mutation Tool	Description
Rename Symbol	Rename operation for types, methods, etc.
Format Document	Format a C# source file using the solution's settings

Public MCP API

These tool descriptions are written as routing triggers. Use them to help an agent decide which tool to call based on the user's intent.

load_solution

Use this tool when you need to start working with a .NET solution and no solution has been loaded yet. This must be the first tool called in a session before any code analysis or navigation tools can be used. The result now includes a readiness state so fresh or detached worktrees can be reported as degraded_missing_artifacts or degraded_restore_recommended instead of leaving users to infer that from diagnostics alone.

Parameters:

• solutionHintPath (optional): Absolute path to a .sln file, or to a directory used as the recursive discovery root for .sln/.slnx files. If omitted, the tool will auto-detect from the current workspace.

understand_codebase

Use this tool when you need a quick overview of the codebase structure at the start of a session. It returns the project structure with dependency relationships and identifies hotspots from hand-written source by default so generated/intermediate artifacts do not dominate the initial view.

Parameters:

• profile (optional): Analysis depth. quick for fast results, standard for balanced output, deep for thorough analysis. Defaults to standard.

list_dependencies

Use this tool when you need to understand how projects relate to each other within a solution. It shows the dependency graph between projects, indicating which projects depend on which others. For automation, prefer projectPath as the stable selector; projectId is snapshot-local to the active workspace snapshot.

Parameters:

• projectPath (optional): Exact path to a project file (.csproj). This is the recommended stable selector for automation. Specify only one of projectPath, projectName, or projectId.
• projectName (optional): Name of a project. Specify only one of projectPath, projectName, or projectId.
• projectId (optional): Project identifier from the current load_solution workspace snapshot. It is snapshot-local and can change after reload, so prefer projectPath for automation. Specify only one of projectPath, projectName, or projectId.
• direction (optional): Which direction of dependencies to return. outgoing shows what the selected project depends on. incoming shows what depends on the selected project. both returns both directions. Defaults to both.

list_types

Use this tool when you need to list types declared in a specific loaded project. It is useful for project-scoped discovery and for finding type symbols by name before calling tools like list_members, resolve_symbol, or get_type_hierarchy. For automation, prefer projectPath as the stable selector; projectId is snapshot-local to the active workspace snapshot. Results prefer handwritten declarations by default, can optionally include XML documentation summaries for returned entries, and report source bias, completeness, and degraded discovery hints.

Parameters:

• projectPath (optional): Exact path to a project file (.csproj). This is the recommended stable selector for automation. Specify only one of projectPath, projectName, or projectId.
• projectName (optional): Name of a project. Specify only one of projectPath, projectName, or projectId.
• projectId (optional): Project identifier from the current load_solution workspace snapshot. It is snapshot-local and can change after reload, so prefer projectPath for automation. Specify only one of projectPath, projectName, or projectId.
• namespacePrefix (optional): Filter to only types in namespaces starting with this prefix.
• kind (optional): Filter by type kind: class, record, interface, enum, or struct.
• accessibility (optional): Filter by accessibility: public, internal, protected, private, protected_internal, or private_protected.
• includeSummary (optional): When true, includes XML documentation summaries for returned type entries when available. Defaults to false.
• limit (optional): Maximum number of results to return. Defaults to 100, maximum 500.
• offset (optional): Number of results to skip for pagination. Defaults to 0.

list_members

Use this tool when you need to inspect the members declared by a specific type. It returns methods, properties, fields, events, and constructors, and supports filtering by kind, accessibility, binding, inheritance, and pagination so you can keep results focused.

Parameters:

• typeSymbolId (optional): The stable symbol ID of a type, obtained from list_types. Provide this OR path+line+column.
• path (optional): Path to a source file. Provide this together with line and column instead of typeSymbolId.
• line (optional): Line number (1-based) pointing to a type in the source file.
• column (optional): Column number (1-based) pointing to a type in the source file.
• kind (optional): Filter by member kind: method, property, field, event, or ctor.
• accessibility (optional): Filter by accessibility: public, internal, protected, private, protected_internal, or private_protected.
• binding (optional): Filter by binding type: static or instance.
• includeInherited (optional): When true, includes members from base classes. Defaults to false.
• limit (optional): Maximum number of results to return. Defaults to 100, maximum 500.
• offset (optional): Number of results to skip for pagination. Defaults to 0.

resolve_symbol

Use this tool when you have a source position (path + line + column), a qualified symbol name, or an existing symbolId and need the stable symbolId plus declaration location used by other navigation tools. This is often the first step before calling explain_symbol, trace_call_flow, find_callers, or find_usages. Qualified-name lookup can search the whole loaded solution, but projectPath is the preferred stable disambiguator for automation.

Parameters:

• symbolId (optional): An existing symbol ID to look up. Provide this OR path+line+column OR qualifiedName.
• path (optional): Path to a source file. Provide this together with line and column instead of symbolId or qualifiedName.
• line (optional): Line number (1-based) in the source file.
• column (optional): Column number (1-based) in the source file.
• qualifiedName (optional): A fully qualified or short type/member name (e.g., System.String, MyNamespace.MyType.MyMethod, or MyMethod). Provide this instead of symbolId or path+line+column.
• projectPath (optional): Optional project scope for qualifiedName lookup — path to a project that contains the symbol. This is the preferred stable selector for automation.
• projectName (optional): Optional project scope for qualifiedName lookup — name of a project that contains the symbol. Use this to narrow ambiguous matches.
• projectId (optional): Optional project scope for qualifiedName lookup — project ID from the current load_solution workspace snapshot. It is snapshot-local and can change after reload, so prefer projectPath when you need a durable selector.

resolve_symbols

Use this tool when you need to resolve multiple symbols in one round-trip. Each entry reuses resolve_symbol semantics, including symbolId, source position, qualifiedName lookup, project scoping, readable symbol references, and structured ambiguity results.

Parameters:

• entries (required): The symbols to resolve. Each entry supports the same selector modes as resolve_symbol: symbolId, path+line+column, or qualifiedName with optional project scoping.

explain_symbol

Use this tool when you need to understand what a specific symbol (type, method, property, field, etc.) does, what its signature looks like, where it is used in the codebase, and what XML documentation it already exposes. It provides a human-readable explanation along with impact hints showing areas with high reference density.

Parameters:

• symbolId (optional): The stable symbol ID, obtained from resolve_symbol, list_types, or list_members. Provide this OR path+line+column.
• path (optional): Path to a source file. Provide this together with line and column instead of symbolId.
• line (optional): Line number (1-based) pointing to the symbol in the source file.
• column (optional): Column number (1-based) pointing to the symbol in the source file.

The result includes a documentation field with:

• summary: The XML summary comment
• returns: The return value description (for methods)
• parameters: List of parameter names and descriptions (for methods)

trace_call_flow

Use this tool when you need to understand how code flows through your system — either finding what calls a specific symbol (upstream) or what a symbol calls (downstream). This is essential for debugging, impact analysis, and understanding architectural patterns. Results prefer hand-written source by default so generated/intermediate call edges do not overwhelm interactive traces, and transition labels now degrade explicitly to unresolved_project/project_inference_degraded when attribution is uncertain. Set includePossibleTargets=true to receive a deliberate possible-runtime-target edge set for uncertain polymorphic dispatch.

Parameters:

• symbolId (optional): The stable symbol ID, obtained from resolve_symbol, list_types, or list_members. Provide this OR path+line+column.
• path (optional): Path to a source file. Provide this together with line and column instead of symbolId.
• line (optional): Line number (1-based) pointing to the symbol in the source file.
• column (optional): Column number (1-based) pointing to the symbol in the source file.
• direction (optional): Which direction to trace. upstream finds callers (who uses this). downstream finds callees (what this calls). both returns both directions. Defaults to both.
• depth (optional): How many levels of the call chain to traverse. Defaults to 2. Use larger values for deeper analysis. null behaves the same as omitting the parameter.
• includePossibleTargets (optional): When true, also returns possible-runtime-target edges for uncertain interface or polymorphic dispatch. Direct static edges remain separate in the main edge list.

find_callers

Use this tool when you need only the immediate direct upstream callers of a symbol. This is a focused wrapper around call-flow tracing and does not traverse beyond one caller level.

Parameters:

• symbolId (optional): The stable symbol ID, obtained from resolve_symbol, list_types, or list_members, for the symbol whose immediate direct callers you want to inspect.

find_callees

Use this tool when you need only the immediate direct downstream callees of a symbol. This is a focused wrapper around call-flow tracing and does not traverse beyond one callee level.

Parameters:

• symbolId (optional): The stable symbol ID, obtained from resolve_symbol, list_types, or list_members, for the symbol whose immediate direct callees you want to inspect.

find_usages

Use this tool when you need to find source-code references to a specific symbol across a document, project, or the entire solution. This is critical before refactoring or modifying a symbol to understand its static impact, but it may not include dynamic, reflection-based, or string-based usages.

Parameters:

• symbolId (required): The stable symbol ID, obtained from resolve_symbol, list_types, or list_members.
• scope (optional): The search scope. project searches only within the containing project. solution searches the entire solution. Defaults to solution.
• path (optional): Required when scope=document: the file path to search within.

find_implementations

Use this tool when you need to find concrete implementations of an interface or abstract type, and overrides or implementations of abstract/virtual members. This is essential for understanding static polymorphic targets in the loaded solution before refactoring or changing a contract.

Parameters:

• symbolId (required): The stable symbol ID of an interface, abstract type, or abstract/virtual member, obtained from resolve_symbol, list_types, or list_members.

get_type_hierarchy

Use this tool when you need to inspect a type's inheritance relationships: base types, implemented interfaces, and derived types. Use includeTransitive=false for immediate parents and children only, or true to expand the full transitive hierarchy.

Parameters:

• symbolId (required): The stable symbol ID of a type, obtained from resolve_symbol, list_types, or list_members. Must resolve to a type (class, interface, enum, struct, or record).
• includeTransitive (optional): When true (default), includes all transitive base types and all derived types. When false, returns only immediate parents and children.
• maxDerived (optional): Maximum number of derived types to return. Defaults to 200. Higher values may impact performance.

find_codesmells

Use this tool when you need to check a specific file for potential code quality issues. It runs Roslyn-based static analysis to detect common problems such as dead code, performance anti-patterns, naming violations, and other code smells identified by Roslynator analyzers. Optional filters operate on stable normalized risk levels (low, review_required, high, info) and categories (analyzer, correctness, design, maintainability, performance, style) in deterministic stream order. reviewMode=conservative favors stronger review signals over low-noise style and trivia suggestions.

Parameters:

• path (required): Path to the source file to analyze. The file must exist in the currently loaded solution.
• maxFindings (optional): Maximum number of accepted findings to return. When provided, discovery stops as soon as this many matching findings are found.
• riskLevels (optional): Accepted risk levels to include. Use normalized result values such as low, review_required, high, or info.
• categories (optional): Accepted categories to include. Use normalized values: analyzer, correctness, design, maintainability, performance, or style. When omitted or empty, all categories are included.
• reviewMode (optional): Review ranking mode. Use 'default' for the existing stream or 'conservative' to suppress lightweight style/trivia noise when stronger issues are present.

rename_symbol

Use this tool when you need to rename a symbol (type, method, property, field, parameter, local variable, etc.) across the entire solution. This performs a safe refactoring that updates all references to the symbol. Returns the list of changed files.

Parameters:

• symbolId (required): The symbol ID of the symbol to rename. Use 'resolve_symbol' to obtain this if needed.
• newName (required): The new name for the symbol. Must be a valid C# identifier and should not conflict with existing symbols in the same scope.

format_document

Use this tool when you need to format exactly one C# source file in the loaded solution using the solution's current formatting and style settings. Returns whether formatting changes were applied and persisted.

Parameters:

• path (required): The path to the C# source file to format. The file must be part of the currently loaded solution.