Objectives: To explore the feasibility of a newly developed model of stratified exercise therapy in primary care for patients with knee osteoarthritis (OA). Design: Mixed method design (process, outcome and qualitative evaluation). Setting: Six physical therapy practices in primary care around Amsterdam. Participants: Fifty eligible patients with knee OA, visiting one of the participating physical therapists (PTs). Intervention: Patients were allocated to a subgroup based on a simple stratification tool and received subgroup-specific, protocolized, 4-month, exercise therapy. Main outcome measures: Feasibility of this model of stratified exercise therapy was explored by multiple process parameters, outcome measures (physical functioning and knee pain; at baseline and 4-months follow-up) and experiences from patients and PTs. Results: From 97 potentially eligible patients, fifty patients were included and allocated to the ‘high muscle strength subgroup’ (n = 17), ‘depression subgroup’ (n = 4), ‘obesity subgroup’ (n = 6) or ‘low muscle strength subgroup’ (n = 23). Three patients dropped out during the study period. PTs provided relatively low numbers of sessions (on average 10 sessions), although exceedance of the recommended maximum number of sessions did occur frequently. We found clinically relevant improvements on physical functioning and knee pain (P < 0.001 for both) for the total group. In general, the model of stratified exercise therapy was considered to be easily applicable and of added value for daily practice. Conclusions: Our model of stratified exercise therapy seems to be feasible in primary care, although a number of limitations were reported. Future research should determine the (cost-)effectiveness of an adapted model, compared to usual, non-stratified exercise therapy.