Google has made the “URL Context” feature in its Gemini large-language-model API generally available, allowing developers to pass one or more URLs—covering webpages, PDFs, images and other files—directly in a prompt. The model fetches the referenced material in real time and integrates it into its response, with users charged only for the additional tokens processed rather than a separate tool fee, according to product documentation released on 18 Aug 2025. The move is aimed at simplifying retrieval-augmented generation workflows and trimming development costs as Google seeks to widen adoption of Gemini among enterprise and consumer-app builders. The upgrade follows recent additions such as native image generation and larger context windows, and intensifies competition with rival AI platforms offering built-in web browsing capabilities. On the same day, Anthropic introduced a Usage and Cost API that lets developers monitor real-time token consumption and spending for its Claude models. The feature, long requested by customers, is designed to help teams optimize prompts and agent architectures amid rapidly scaling generative-AI workloads.