Digital technology is positioned as a potential solution to improving access, experience, and outcomes of psychological therapies for psychosis. Digital solutions need to be fit for purpose and tailored to context to deliver real world benefits. To address this, co-production is often used, where stakeholder involvement informs intervention development. However, co-production in clinical research tends to limit involvement to refining previously identified solutions to known problems. This is not an optimal approach to innovation and risks maintaining inequities. An alternative is inclusive co-design, where the needs of a diverse range of people are collaboratively explored using ethnography, and solutions to address these iteratively developed through user testing. In healthcare, we propose an evidence-based approach to co-design (‘hybrid waterfall-agile’) is required. This is because ‘agile’ exploration of needs and solutions is necessarily constrained by clinical guidelines and regulatory requirements (the ‘waterfall’). This paper provides an overview of evidence-based co-design. We use the example of SloMo, a blended digital therapy for paranoia. We describe our transdisciplinary team collaboration and how this facilitates inclusive lived experience involvement. Our therapy development method is outlined, illustrated by reflections from lived experience team members. Iterative divergent (‘zooming out’) and convergent (‘honing in’) cycles are used to co-design therapy functionality, aesthetics, interactions, and content, supported by stakeholder engagement. We conclude by reflecting on common challenges including sustaining lived experience involvement, adherence to evidence base, regulatory compliance, funding, and project management. Recommendations for navigating these obstacles are provided, with the aim of encouraging innovation in mental healthcare for psychosis.