Abstract Background Injections targeting the occipital nerve are used to reduce headache attacks and abort cluster bouts in cluster headache patients. There is no widely accepted agreement over the optimal technique of injection, type and doses of steroids and/or anesthetics to use, as well as injection regimens. The aim of this study was to verify the effectiveness and safety of greater occipital nerve long-acting steroid injections in the management of episodic and chronic cluster headache. Methods We conducted a prospective observational cohort study on episodic (ECH) and chronic cluster headache patients (CCH). ECH were included in the study at the beginning of a cluster period. Three injections with 60 mg methylprednisolone were performed on alternate days. We registered the frequency and intensity of attacks three days before and 3, 7 and 30 days after the treatment, the latency of cluster relapse, adverse events, scores evaluating anxiety (Zung scale), depression (Beck’s Depression Scale) and quality of life (Disability Assessment Schedule II, 12-Item Self-Administered Version). Primary outcome was the interruption of the cluster after the three injections. Responders conducted a follow-up period of 12 months. Results We enrolled 60 patients, 47 with ECH and 13 with CCH. We observed a complete response in 47.8% (22/46) of episodic and 33.3% (4/12) of chronic patients. Moreover, a partial response (reduction of at least 50% of attacks) was obtained in further 10.8% (5/46) of episodic and in 33.3% (4/12) of chronic patients at 1 month. Median pain-free period was of 3 months for CCH responders. Only mild adverse events were reported in 38.3% (23/58) cases. Conclusions We suggest three greater occipital nerve injections of 60 mg methylprednisolone on alternate days as useful therapy in episodic and chronic cluster headache. This leads to a long pain-free period in chronic forms. Adverse effects are mild and support its use as first choice. Trial registration The study was inserted in AIFA observational studies register.