PRISM Open-Source Tool Reveals Real-Time Carbon Cost of AI for Developers | CarbonCredit.io

A new open-source tool called PRISM, developed by ustwo and the University of Bristol, gives software developers real-time estimates of the carbon emissions generated by their AI use during coding. The plugin integrates into development environments and provides directional carbon impact data for each AI query, focusing on inference the day-to-day use of AI models that is becoming a major energy drain. PRISM uses a token-to-energy-to-carbon methodology informed by the Green Software Foundation and academic research. It aims to shift sustainability left into the code-writing stage, making carbon impact a visible metric alongside performance and cost. The tool is freely available and open-source, inviting community collaboration to refine measurement standards. For carbon markets and climate tech professionals, PRISM addresses a growing blind spot: the energy footprint of AI inference. While training emissions get attention, inference emissions are rising fast. Tools like this could help companies align AI development with net-zero goals and make granular carbon data actionable at the developer level.