mcp-server-tree-sitter

Instructions

Implementation Reference

• src/mcp_server_tree_sitter/tools/registration.py:496-513 (registration)

  Registration of the 'get_dependencies' MCP tool using @mcp_server.tool() decorator. This is the entry point handler that receives tool arguments and delegates to the analysis module.
  def get_dependencies(project: str, file_path: str) -> Dict[str, List[str]]: """Find dependencies of a file. Args: project: Project name file_path: Path to the file Returns: Dictionary of imports/includes """ from ..tools.analysis import find_dependencies return find_dependencies( project_registry.get_project(project), file_path, language_registry, )
• src/mcp_server_tree_sitter/tools/registration.py:496-513 (handler)

  Handler function for the 'get_dependencies' tool, which extracts dependencies from a file using Tree-sitter queries via the analysis module.
  def get_dependencies(project: str, file_path: str) -> Dict[str, List[str]]: """Find dependencies of a file. Args: project: Project name file_path: Path to the file Returns: Dictionary of imports/includes """ from ..tools.analysis import find_dependencies return find_dependencies( project_registry.get_project(project), file_path, language_registry, )
• src/mcp_server_tree_sitter/tools/analysis.py:594-794 (helper)

  Core implementation of dependency extraction using Tree-sitter queries to find import/include statements in the specified file, categorizing and deduplicating them across supported languages.
  def find_dependencies( project: Any, file_path: str, language_registry: Any, ) -> Dict[str, List[str]]: """ Find dependencies of a file. Args: project: Project object file_path: Path to the file relative to project root language_registry: Language registry object Returns: Dictionary of dependencies (imports, includes, etc.) """ abs_path = project.get_file_path(file_path) try: validate_file_access(abs_path, project.root_path) except SecurityError as e: raise SecurityError(f"Access denied: {e}") from e language = language_registry.language_for_file(file_path) if not language: raise ValueError(f"Could not detect language for {file_path}") # Get the appropriate query for imports query_string = get_query_template(language, "imports") if not query_string: raise ValueError(f"Import query not available for {language}") # Parse file and extract imports try: # Get language object language_obj = language_registry.get_language(language) safe_lang = ensure_language(language_obj) # Parse with cached tree tree, source_bytes = parse_with_cached_tree(abs_path, language, safe_lang) # Execute query query = safe_lang.query(query_string) matches = query.captures(tree.root_node) # Organize imports by type imports: Dict[str, List[str]] = defaultdict(list) # Track additional import information to handle aliased imports module_imports: Set[str] = set() # Helper function to process an import node def process_import_node(node: Any, capture_name: str) -> None: try: safe_node = ensure_node(node) text = get_node_text(safe_node, source_bytes) # Determine the import category if capture_name.startswith("import."): category = capture_name.split(".", 1)[1] else: category = "import" # Ensure we're adding a string to the list text_str = text.decode("utf-8") if isinstance(text, bytes) else text imports[category].append(text_str) # Add to module_imports for tracking all imported modules if category == "from": # Handle 'from X import Y' cases parts = text_str.split() if parts: module_part = parts[0].strip() module_imports.add(module_part) elif category == "module": # Handle 'import X' cases text_str = text_str.strip() module_imports.add(text_str) elif category == "alias": # Handle explicitly captured aliases from 'from X import Y as Z' cases # The module itself will be captured separately via the 'from' capture pass elif category == "item" and text: # For individual imported items, make sure to add the module name if it exists if hasattr(safe_node, "parent") and safe_node.parent: parent_node = safe_node.parent # The import_from_statement node # Find the module_name node for child in parent_node.children: if ( hasattr(child, "type") and child.type == "dotted_name" and child != safe_node and hasattr(child, "text") ): module_name_text = get_node_text(child, source_bytes) module_name_str = ( module_name_text.decode("utf-8") if isinstance(module_name_text, bytes) else module_name_text ) module_imports.add(module_name_str) break elif "import" in text_str: # Fallback for raw import statements parts = text_str.split() if len(parts) > 1 and parts[0] == "from": # Handle 'from datetime import datetime as dt' case part = parts[1].strip() module_imports.add(str(part)) elif "from" in text_str and "import" in text_str: # Another way to handle 'from X import Y' patterns # text_str is already properly decoded from_parts = text_str.split("from", 1)[1].split("import", 1) if len(from_parts) > 0: module_name = from_parts[0].strip() module_imports.add(module_name) elif parts[0] == "import": for module in " ".join(parts[1:]).split(","): module = module.strip().split(" as ")[0].strip() module_imports.add(module) except Exception: # Skip problematic nodes pass # Handle different return formats from query.captures() if isinstance(matches, dict): # Dictionary format: {capture_name: [node1, node2, ...], ...} for capture_name, nodes in matches.items(): for node in nodes: process_import_node(node, capture_name) else: # List format: [(node1, capture_name1), (node2, capture_name2), ...] for match in matches: # Handle different return types from query.captures() if isinstance(match, tuple) and len(match) == 2: # Direct tuple unpacking node, capture_name = match elif hasattr(match, "node") and hasattr(match, "capture_name"): # Object with node and capture_name attributes node, capture_name = match.node, match.capture_name elif isinstance(match, dict) and "node" in match and "capture" in match: # Dictionary with node and capture keys node, capture_name = match["node"], match["capture"] else: # Skip if format is unknown continue process_import_node(node, capture_name) # Add all detected modules to the result if module_imports: # Convert module_imports Set[str] to List[str] module_list = list(module_imports) imports["module"] = list(set(imports.get("module", []) + module_list)) # For Python, specifically check for aliased imports if language == "python": # Look for aliased imports directly aliased_query_string = "(aliased_import) @alias" aliased_query = safe_lang.query(aliased_query_string) aliased_matches = aliased_query.captures(tree.root_node) # Process aliased imports for match in aliased_matches: # Initialize variables aliased_node: Optional[Any] = None # We're not using aliased_capture_name but need to unpack it _: str = "" # Handle different return types if isinstance(match, tuple) and len(match) == 2: aliased_node, _ = match elif hasattr(match, "node") and hasattr(match, "capture_name"): aliased_node, _ = match.node, match.capture_name elif isinstance(match, dict) and "node" in match and "capture" in match: aliased_node, _ = match["node"], match["capture"] else: continue # Extract module name from parent if aliased_node is not None and aliased_node.parent and aliased_node.parent.parent: for child in aliased_node.parent.parent.children: if hasattr(child, "type") and child.type == "dotted_name": module_name_text = get_node_text(child, source_bytes) if module_name_text: module_name_str = ( module_name_text.decode("utf-8") if isinstance(module_name_text, bytes) else module_name_text ) module_imports.add(module_name_str) break # Update the module list with any new module imports if module_imports: module_list = list(module_imports) imports["module"] = list(set(imports.get("module", []) + module_list)) return dict(imports)