Strava Just Gave Claude the Keys to Your Training Data
Strava's new Claude MCP gives subscribers live AI access to training data — but only through one platform. Here's what it does and what it costs.
Strava just opened a direct pipeline between your training data and Claude, Anthropic's AI assistant. Starting June 1, subscribers can connect via a Model Context Protocol (MCP) that gives Claude live, read-only access to your full activity history — heart rate streams, GPS tracks, power files, the works. You ask natural-language questions; Claude answers from your actual data, not a CSV you uploaded last month.
This matters because Strava has spent the last two years locking down third-party API access. They've made it progressively harder for developers to build tools on top of your data, even when you wanted them to. Now they're handing the keys to one AI platform while keeping everyone else out. It's selective openness, and whether that's user-driven demand or a calculated partnership depends on how charitable you feel about corporate strategy.
The practical upside is real. Before this, querying your training history through an LLM meant exporting a bulk file, pasting it into Claude or ChatGPT, and repeating the whole process every time you logged another week. The MCP persists across sessions and pulls live data, so you can ask "Are my easy runs actually easy?" or "Did that strength block improve my cycling power?" without manually updating a spreadsheet. You also get access to subscription features — fitness trends, training load analytics — outside the Strava app for the first time.
The connector is read-only and revocable from your Strava settings. You'll need a paid Strava subscription (standard, family, student, or bundled with Runna). Setup works through Claude's web interface, desktop app, or CLI. Rate limits apply, though Strava hasn't published the exact thresholds. And Claude's interpretations won't always match what you see in the Strava app — different models, different math.
If you've been exporting .fit files and writing Python scripts to answer questions Strava won't surface, this is faster. If you're skeptical about why Strava chose one AI vendor after years of closing the API gates, you're not wrong to wonder. Either way, it's a subscriber-only feature that makes your data more queryable. Just know you're opting into an AI platform that Strava picked for you.
More info --> Strava
