Agent-based crowd simulation is widely used to assess the safety of buildings and events and represents a powerful tool for supporting evacuation planning and design. Existing models typically differentiate agents only by assigning distributions for free walking speed or torso size or variation in route choice, thereby capturing limited variability in human characteristics. In this contribution, we extend the commercial agent-based tool crowd:it [1] to explicitly represent assisted evacuation scenarios with heterogeneous populations. Building on the Optimal Steps Model [2] and its implementation in crowd:it, we introduce two new agent types, Immobile and Rescuer, to capture evacuation situations in which not all occupants can evacuate autonomously. Immobile agents are unable to move independently and require the presence of a nearby Rescuer to be evacuated, while Rescuer agents repeatedly re-enter the building to search for and escort remaining Immobile agents. The model allows configuring this behaviour through heuristics, such as restricting rescues to specific spatial areas or prioritising Immobile agents within a given distance threshold. This extension enables a more realistic representation of assisted evacuations in facilities such as hospitals and retirement homes, where a significant fraction of occupants depends on staff or other helpers. We will present the approach using a real-world case study and demonstrate how the extended model can support the design and evaluation of fast and efficient evacuation strategies for vulnerable user groups.