The Green's Function Model Intercomparison Project (GFMIP) protocol is introduced to standardize the creation of atmospheric Green's functions for model intercomparison. This protocol, developed using insights from past studies and new sensitivity tests, aims to better understand the atmospheric response to surface temperature changes, particularly the "pattern effect" where the response depends on the spatial distribution of warming. Green's functions reconstruct historical temperature responses but may be affected by nonlinearities in atmospheric responses to different warming patterns. The protocol balances scientific utility with computational constraints, using the HadAM3 model and other simulations to explore nonlinearities in atmospheric responses, such as asymmetrical responses to warming and cooling patches and changes in response magnitude and size. The protocol includes control simulations, patch simulations, and diagnostic simulations, with boundary conditions available at https://gfmip.org. Sensitivity tests with HadAM3 show that using different base states can affect the skill of Green's function reconstructions, highlighting the importance of consistent boundary conditions. The protocol aims to ensure that differences in derivatives are due to model differences rather than experimental setup. The protocol includes parameters for control, patch, and diagnostic simulations, with optional simulations for different perturbation types. The goal is to improve understanding of atmospheric responses, including radiative feedbacks, circulation, cloudiness, and precipitation, by comparing model results across different warming patterns.The Green's Function Model Intercomparison Project (GFMIP) protocol is introduced to standardize the creation of atmospheric Green's functions for model intercomparison. This protocol, developed using insights from past studies and new sensitivity tests, aims to better understand the atmospheric response to surface temperature changes, particularly the "pattern effect" where the response depends on the spatial distribution of warming. Green's functions reconstruct historical temperature responses but may be affected by nonlinearities in atmospheric responses to different warming patterns. The protocol balances scientific utility with computational constraints, using the HadAM3 model and other simulations to explore nonlinearities in atmospheric responses, such as asymmetrical responses to warming and cooling patches and changes in response magnitude and size. The protocol includes control simulations, patch simulations, and diagnostic simulations, with boundary conditions available at https://gfmip.org. Sensitivity tests with HadAM3 show that using different base states can affect the skill of Green's function reconstructions, highlighting the importance of consistent boundary conditions. The protocol aims to ensure that differences in derivatives are due to model differences rather than experimental setup. The protocol includes parameters for control, patch, and diagnostic simulations, with optional simulations for different perturbation types. The goal is to improve understanding of atmospheric responses, including radiative feedbacks, circulation, cloudiness, and precipitation, by comparing model results across different warming patterns.