Background Housing First does not, on average, reduce criminal justice involvement. This analysis aims to test whether the overall absence of an impact is due to intervention effect heterogeneity as a function of the pattern of lifetime criminal justice involvement, identified through latent class analysis conducted through earlier work. Methods This analysis relied on data from the Montréal, Toronto and Vancouver sites of the Canadian At Home/Chez Soi randomized controlled trial, merged with administrative records of lifetime criminal charges (N = 1,321). Negative binomial models with interaction terms were used to estimate the impact of Housing First, in comparison to treatment as usual, on violent charges, acquisitive charges (e.g., theft, sex work), and administration of justice charges (e.g., breach of probation), for each pre-identified profile. Results Participants with past criminal justice involvement associated with a chronic history of homelessness or with criminalized substance use experienced a decrease in violent charges as a result of Housing First, whereas those with no or little past criminal justice involvement experienced a marginal increase. Housing First did not affect acquisitive or administration of justice charges, regardless of profile. Conclusions Findings suggest that integrating criminological or forensic mental health tools, knowledge and approaches into the multidisciplinary teams that support Housing First service users may be an effective solution, so that all aspects of their recovery, including potential criminogenic needs, are addressed. Future research should focus on the feasibility and effectiveness of such adjunct interventions.