Nodes of the control-flow graph. Every node is either a conditional node, switch node, leaf node, or expression node.
Edges of the control-flow graph, representing control-flow between nodes.
Expression nodes, similar in nature to abstract syntax trees (AST) for expressions. Each node may contain a read expression (right-hand side) and write expression (left-hand side).
Memory items, which are conceptual objects that represent some abstract location in memory. Not every possible memory location is represented by a memory item. For example, there are no memory items for individual array elements, only for the whole array. A memory item is an object that allows value tracking across multiple nodes. Expressions cannot be used to do this because every node has a unique instance of an expression. However, it is possible to extract memory items from two expressions and see if they are the same.
Semantic information, which describes the program data and types. This information is somewhat limited compared to semantic information that can be derived using the Code Style Analysis API.
Value constraints, which are constraints on integer values inferred by the Klocwork engine. You can query the Klocwork C/C++ Path API Reference PDF about whether values for certain memory items lie within some specific range.

The MIR is built separately for each function and class method.