Abstract Background Associations between cannabis use and psychotic outcomes are consistently reported. Increased innate immune markers have also been suggested as risk factors for psychosis; nevertheless, no study has ever tested whether augmented blood inflammation could be a possible biological mechanism underlying the association between cannabis use and psychosis. We investigated: i) which patterns of peripheral blood cytokines (innate and adaptive immune markers) would be associated with the strongest effect on odds of psychosis; and ii) whether the contribution of cannabis use (lifetime use: yes/no; frequency of use: non-users, less than daily use; daily use) on risk of psychosis would vary between subgroups with low- or high-inflammation. Methods The Schizophrenia and Other Psychoses Translational Research: Environment and Molecular Biology (STREAM) is an epidemiological and case-sibling-control study, conducted in Ribeirão Preto catchment area (São Paulo, Brazil) between April 2012 and March 2015, which integrates the international multicentre consortium EU-GEI. We recruited 153 first-episode psychosis patients and 256 community-based controls aged between 16–64 years. Participants answered the Cannabis Experience Questionnaire and plasma cytokines (IL-1β, IL-6, TNF-α, IFN-γ, IL-4, IL-10, TGF-β) were measured by Multiplex. We firstly investigated which patterns of peripheral blood innate and adaptive immune markers would contribute to the strongest effect on the odds of psychosis. To do that, the sample was divided into quartiles according to cytokine’s percentile distribution (reference group: minimum value – 25th percentile; lower group: 25th – 50th percentile; middle group: 50th – 75th percentile; and upper group: 75th percentile – maximum value). We next investigated whether the effects of the different patterns of cannabis use (lifetime and frequency of use) on the risk of psychosis would differ between subgroups classified as low- or high-inflammatory, using the median (50th percentile) of each cytokine as the cut-off value. All the binary logistic regression analyses were correct for the effects of confounders (gender, age, ethnicity, years of education, body mass index, tobacco smoking, and recreational drugs). Results More than 60% of the patients were classified as high-inflammatory, whereas in the control group, around the same percentage was classified as low-inflammatory. After adjusting for confounders, we observed a dose-response relationship between the percentile’s distribution of IL-10, IL-6, TNF-α, TGF-β and risk of psychosis, with participants classified in the upper percentile having the highest odds ratio when compared with the reference group. The adjusted odds ratio (adj ORs) and 95% CIs range for each cytokine were the following: IL-10 from 4.68 (2.02–10.85) to 11.86 (5.01–28.11), IL-6 from 3.74 (1.61–8.69) to 9.62 (4.15–22.31); TNF-α from 2.18 (1.01–4.73) to 6.88 (3.12–15.17); TGF-β from 2.17 (1.01–4.67) to 3.20 (1.50–6.80). Lifetime cannabis use only increased risk of psychosis in the high-inflammatory subgroup (adj ORs, 95% CIs: IL10: 4.66, 1.62–13.39; IL-6: 2.75, 1.05–7.21). Daily use increased the risk of psychosis even further, and only in the high-inflammatory but not in the low-inflammatory subgroup (adj OR, 95% CIs: IL-10: 14.09, 2.26–87.93; TNF-α: 6.16, 1.28–29.74; IL-6: 4.68, 1.08–20.36), with the exception of TGF-β, for which the effects were seen in the low-subgroup (adj OR, 95% CI: 5.23, 1.17–23.42). Discussion The existence of distinct inflammatory profiles indicates possible biological predisposition facilitating immune activation. This would translate into a higher vulnerability to the effects of cannabis use, especially daily use, on risk of psychosis.