Nuanced MCP Server

Instructions

Implementation Reference

• nuanced_mcp_server.py:25-64 (handler)

  The handler function for the 'initialize_graph' tool, including the @mcp.tool() decorator which also serves as registration. This function initializes a CodeGraph for the given repository path, stores it globally, sets it as active, and counts Python files.

  @mcp.tool()
  def initialize_graph(repo_path: str) -> str:
      """Initialize a code graph for the given repository path.
  
      Args:
          repo_path: Path to the repository to analyze
  
      Returns:
          Success message with information about the initialized graph
      """
      global _code_graphs, _active_repo
  
      # Check if path exists
      if not os.path.exists(repo_path):
          return f"Error: Path '{repo_path}' does not exist"
      
      # Use absolute path as key
      abs_path = os.path.abspath(repo_path)
  
      try:
          result = CodeGraph.init(repo_path)
  
          if result.errors:
              error_messages = [str(error) for error in result.errors]
              return f"Errors initializing code graph:\n" + "\n".join(error_messages)
  
          _code_graphs[abs_path] = result.code_graph
          _active_repo = abs_path
  
          # Count Python files
          py_files = glob.glob(f"{repo_path}/**/*.py", recursive=True)
  
          return (
              f"Successfully initialized code graph for {repo_path}.\n"
              f"Repository contains {len(py_files)} Python files.\n"
              f"This is now the active repository."
          )
      except Exception as e:
          return f"Error initializing code graph: {str(e)}"

• nuanced_mcp_server.py:25-25 (registration)

  The @mcp.tool() decorator registers the initialize_graph function as an MCP tool.

  @mcp.tool()