Multi-agent epistemic planning is a subfield of artificial intelligence that allows modelling for domains with multiple agents and their beliefs. In this framework, each agent can reason using their beliefs about their environment, other agents, and other agents’ beliefs. Existing planners for this formalism have explored notions from the psychological theory of mind, including nested beliefs and even deception. However, advances in epistemic planning have mainly developed independently of these frameworks. We bridge this gap by applying the strengths of classical planning to bear on a novel interpretation of reasoning. We introduce RP-M*P, an extension of a state-of-the-art epistemic planner (RP-MEP) that compiles to classical planning and augments it with nested modalities for desire and intention. Our novel planner features a new syntax for specifying planning problems with agent belief, desire, and intention modalities, theory of mind across agents, a formalism for intending actions, and customizable ancillary effects. These customized ancillary effects enable flexible domain-specific agent behaviour, such as custom psychological notions of modalities (for example, a “skeptical” agent that refuses to adopt another’s beliefs). Our framework of beliefs, desires, and intentions is loosely inspired by the well-known BDI frameworks, but we focus on customizability and allowing for future extensions, e.g. emotional states. We assessed the planner’s performance against the traditional “Grapevine” benchmark and three new handcrafted domains, assessing compiled domain size, compilation time, and planning time. Results demonstrate promising planning times, and the domains exhibit a truly rich setting for multi-agent interaction.