IDA Pro MCP

Instructions

Implementation Reference

• src/ida_pro_mcp/ida_mcp/api_core.py:278-308 (handler)

  Handler function implementing the list_globals tool. Collects global variables from IDA's Names(), filters non-function named addresses, applies pagination and pattern filtering.

  @tool
  @idaread
  def list_globals(
      queries: Annotated[
          list[ListQuery] | ListQuery | str,
          "List global variables with optional filtering and pagination",
      ],
  ) -> list[Page[Global]]:
      """List globals"""
      queries = normalize_dict_list(
          queries, lambda s: {"offset": 0, "count": 50, "filter": s}
      )
      all_globals: list[Global] = []
      for addr, name in idautils.Names():
          if not idaapi.get_func(addr) and name is not None:
              all_globals.append(Global(addr=hex(addr), name=name))
  
      results = []
      for query in queries:
          offset = query.get("offset", 0)
          count = query.get("count", 100)
          filter_pattern = query.get("filter", "")
  
          # Treat empty/"*" filter as "all"
          if filter_pattern in ("", "*"):
              filter_pattern = ""
  
          filtered = pattern_filter(all_globals, filter_pattern, "name")
          results.append(paginate(filtered, offset, count))
  
      return results

• src/ida_pro_mcp/ida_mcp/utils.py:230-232 (schema)

  Output item schema: Global TypedDict defining structure of each global variable (address and name). Used in Page[Global].

  class Global(TypedDict):
      addr: str
      name: str

• src/ida_pro_mcp/ida_mcp/utils.py:127-132 (schema)

  Input schema: ListQuery TypedDict for optional filter, offset, and count parameters supporting pagination and filtering.

  class ListQuery(TypedDict, total=False):
      """Pagination query for listing operations"""
  
      filter: Annotated[str, "Optional glob pattern to filter results"]
      offset: Annotated[int, "Starting index (default: 0)"]
      count: Annotated[int, "Maximum number of results (default: 50, 0 for all)"]

• src/ida_pro_mcp/ida_mcp/utils.py:357-359 (schema)

  Pagination response wrapper: Page generic TypedDict used for output structure list[Page[Global]].

  class Page(TypedDict, Generic[T]):
      data: list[T]
      next_offset: Optional[int]

• src/ida_pro_mcp/ida_mcp/rpc.py:17-18 (registration)

  The @tool decorator registers functions as MCP tools by adding them to the McpServer instance.

  def tool(func):
      return MCP_SERVER.tool(func)