Developer LisaMarieJosefine released Stratoscope, an open-source menu bar utility for macOS and Windows that monitors local LLM inference metrics in real time. The tool displays prefill and decode token throughput, cache hit rates, context window utilization, and system resource usage across RAM, GPU, and CPU. It currently integrates with the Strata inference engine via a lightweight adapter architecture designed to support additional engines with minimal effort. The project is available on GitHub under the MIT license.