I've wasted way too much time this year managing different API keys for each of my providers, OpenAI, Anthropic, and Google in order to run simple A/B tests between different prompts on different models. Hopscotch solves this problem right away by providing one endpoint to 500+ models and charging me the provider rates rather than adding a markup. The thing I really would like to have now would be latency statistics for each model in the dashboard.