Hi Louis, I'm a long-time Calistree user from Brazil (@handstand_project) and I'd like to pitch a feature that I believe could benefit a lot of your users. Here's how I use the app today. I run a long-term rehab and skill project (handstand, front lever) with documented shoulder and hip issues, and Calistree is my single source of truth — every session, every load, every note. On top of that, I use Claude (Anthropic's AI assistant) as a training advisor: it tracks load progressions across weeks, watches for recurring pain signals in my notes, enforces progression criteria (e.g., two clean sessions before adding weight), and maintains a longitudinal training log. The friction is the bridge between the two: today it's manual PDF exports, one session at a time. The PDFs lose structure (custom exercise labels, units), and doing a monthly trend review means feeding a stack of files into the AI by hand. The idea: a Calistree MCP server. MCP (Model Context Protocol) is an open standard — created by Anthropic and since adopted across the industry, including OpenAI and Google — that lets apps expose data to AI assistants in a structured way. A read-only server with a handful of endpoints would already cover the core use case: list workouts by date range exercise history (sets, reps, load, notes) for a given exercise personal records and streaks With that in place, any user could connect their training history to their AI of choice and ask things like "how has my bench progressed over the last 8 weeks?" or "flag every exercise where my reps dropped at the same load." For self-coached athletes and coaches alike, that turns Calistree into the data layer of their whole training stack — which I'd expect to be a real differentiator and a retention driver, since the value of the logged history compounds over time. I'd be happy to beta test and share my workflow in detail if that's useful. Thanks for the great app, Rodrigo B. Uehara @handstand_project