The conflict between agricultural development and fragile ecosystems in arid regions is inevitable, highlighting the importance of synergizing the relationship between oasis agricultural economies and resource environments. This study focuses on Xinjiang, China, analyzing data from 2001 to 2020 to assess the total carbon emissions and intensity of crop and livestock husbandry. Utilizing the LMDI decomposition method and the Tapio decoupling model, this research examines the influencing factors of agricultural carbon emissions and their decoupling effects. The findings indicate that: (1) The total carbon emissions from agriculture in Xinjiang have steadily increased, with carbon intensity decreasing from 1.24 t per 10,000 yuan in 2001 to 0.24 t per 10,000 yuan in 2020, reflecting a significant downward trend. In terms of the composition of agricultural carbon emissions, the primary sources of emissions in both crop husbandry and livestock husbandry are fertilizers and enteric fermentation from livestock, respectively; (2) Structural and efficiency factors have both moderated the growth of carbon emissions, whereas economic and labor factors have contributed to their increase, with the emissions driven by promoting factors surpassing the reductions achieved through inhibiting factors; (3) The decoupling status of carbon emissions in crop husbandry and livestock husbandry exhibits fluctuations, but shows an overall trend toward strong or weak decoupling. By increasing investment in technology, optimizing industrial structures, and promoting a low-carbon and green transformation of agriculture, this study provides theoretical guidance and empirical evidence for achieving the "dual carbon" goals in arid and semi-arid regions.