Objective: Psychological interventions reduce the impact of psychosis, but widescale implementation is problematic. We tested the feasibility of group acceptance and commitment therapy for Psychosis (G‐ACTp), delivered by frontline staff, and co‐facilitated by service‐user experts‐by‐experience (SU‐EbyE), for service‐users and informal caregivers (ISRCTN: 68540929). We estimated recruitment/retention rates and outcome variability for future evaluation. Methods: Staff and SU‐EbyE facilitators completed 1‐day workshops, then delivered closely supervised G‐ACTp, comprising four sessions (weeks 1–4) and two boosters (10 and 12 weeks). Participants recruited from adult community psychosis services were randomized to receive G‐ACTp immediately or after 12 weeks, completing outcome assessments at 0, 4, and 12 weeks. Service‐use/month was calculated for 1‐year pre‐randomization, weeks 0–12, and 5‐year uncontrolled follow‐up. Results: Of 41 facilitators trained (29 staff, 12 SU‐EbyE), 29 (71%; 17 staff, 12 SU‐EbyE) delivered 18 G‐ACTp courses. Participant refusal rates were low (9% of service‐users [10/112]; 5% of caregivers [4/79]); 60% of those invited to participate attended ≥1 G‐ACTp session (64% of service‐users [39/61]; 56% of caregivers [35/63]). Randomization of facilitators and participants proved problematic and participant follow‐up was incomplete (78% [66/85]; 82% of service‐users [36/44]; 73% of caregivers [30/41]). Effect sizes ranged from very small to large mostly favouring treatment. Service‐use reductions require cautious interpretation, as very few participants incurred costs. Conclusions: Implementation appears feasible for service‐users; for caregivers, retention needs improving. Outcome variability indicated n = 100–300/arm followed up (α = 0.05, 90% power). Methodological limitations' mean replication is needed: identified sources of potential bias may be reduced in a cluster randomized design with sessional outcome completion. Practitioner points: Group acceptance and commitment therapy can be successfully adapted for people with psychosis and their caregivers.Implementation (training and delivery) is possible in routine community mental health care settings.Clinical and economic outcomes are promising, but replication is needed.Recommendations are made for future studies.