fitdocs
An AI-native training log for analysis and planning.
Why fitdocs?
Fitdocs takes your workout files and builds a training log and an analytics database, in the format that LLMs read best. It runs locally on your computer, and works with any AI agent without additional setup. Use it as a standalone training log, or integrate it into a markdown knowledge base you already keep, like an Obsidian vault, a git-tracked wiki, or an LLM-maintained "second brain." All major fitness wearables output .fit files, so you can connect your fitness data from Garmin, Coros, Polar, Stryd, Apple, and more.
Analyze your training with AI¶
Large language models (LLMs) are powerful resources for analyzing and planning training. Fitdocs transforms the workout files from your wearable into an LLM-native format, and includes a powerful analytics database for immediate insight into patterns across years of training. The knowledge store is a local directory of linked markdown files alongside a lightning-fast duckdb analytics database.
Fitdocs runs the numbers, LLMs make the connections¶
LLMs are great at reading, connecting, and explaining. They aren't a reliable calculator. Fitdocs calculates load, zones, splits, benchmarks, and fitness curve without any AI. Your LLM put the stats from your latest workout into context, identifies fitness trends over time, and writes the training plan you describe into the wiki, where it’s matched with