There is growing evidence that climate change poses significant risks to agriculture in many parts of the world, especially to the small-scale farmers in Eastern Africa region because of its high dependence on rainfed agriculture which is highly sensitive to climate variability and change. Despite the availability of overwhelming evidence in support of large negative impacts of climate change on agriculture, uncertainty prevails over the precise nature of projected changes in climate and their effects on agricultural systems, especially at point scale levels, making it difficult to plan and develop appropriate adaptation strategies, programs, and technologies. Under the global Agricultural Model Intercomparison and Improvement Project (AgMIP), teams of researchers from four Eastern Africa countries viz., Ethiopia, Kenya, Tanzania and Uganda have initiated studies to make a comprehensive assessment of climate change impacts on smallholder agriculture and this paper examines the farm level impacts of climate change on maize productivity in Embu county, Kenya. Agriculture in this county is predominantly characterized as smallholder, with farm sizes ranging from 0.9 to 2.5 hectares, dedicated to subsistence with limited marketable surpluses. Maize is the most important staple food crop and is grown by almost all farmers under diverse conditions across the county. The average annual rainfall varies from more than 1200 mm to less than 600 mm and the average annual temperatures varies from about 20⁰C at an altitude of 1500 m to about 24⁰C at 700 m. Most farmers use labor intensive traditional farming methods that include use of local varieties and very low levels of inputs such as fertilizers. Mean grain yields of maize are generally low and vary from 1 t/ha in high potential areas to less than 0.5 t/ha in low potential areas. Crop yields under current and future climates were simulated using crop models APSIM and DSSAT by setting model parameters that reflect farmer management practices and using observed climate data and generated future climates for A2 and B1 emission scenarios based on the Hadcm3 GCM for the mid-century period (2040-2069). On an average, the temperature increased by about 1.89⁰C under the A2 scenario and by about 1.19⁰C under  B1 scenario. The models APSIM and DSSAT differed in simulating maize temperature response. While DSSAT predicted a 15% decline in yields under all test environments and management conditions, APSIM predicted an increase in yield in some environments especially those under low input systems with less than 25 kg N/ha. Further analysis of these changes in maize yield with the Tradeoff Analysis Minimum Data model (TOA-MD) has indicated that with current technology and a changed climate--with increased temperature and rainfall--a mean of 51% of all farmers in all environments will be better off. However, there are differences across environments and across management practices. The paper discusses these differences and their implications for climate change adaptation.