Teacher classroom practices have been identified as the key contributing factor to the low learning outcomes of primary school grades in Kenya. Teacher mentoring has the potential to improve classroom practices and this study sought to determine its effects on the classroom practices of primary school teachers in Kwale County, Kenya. Using One Group Repeated Measures Quasi-Experimental Design, one cohort of 40 teachers in 22 public primary schools was mentored for 20 months, from May 2016 to October 2018. A classroom observation tool was used in data collection during this period. The tool achieved a Cronbach's Alpha reliability coefficient of 0.84, 0.81 and 0.79 in 2016, 2017 and 2018 respectively. Data was analysed using Repeated Measures ANOVA and teacher mentoring had a statistically significant effect on mean classroom practice score at F=6.282, df=2, p=0.003. Significant mean differences were located between 2016 and 2017 in favour of 2017, and between 2016 and 2018 in favour of 2018. There was no significant mean difference between 2017 and 2018. In conclusion, teacher mentoring is effective in improving teacher classroom practices and should be integrated into the formal school programme in Kenya.