If you train with Strava and use Claude, you've probably had the thought: why can't I just ask Claude about my running or cycling data directly? Instead of scrolling through dashboards, you'd just type "Am I overtraining this week?" or "What was my best 5k this month?" and get a real answer based on your actual activities.

The good news: you can. The Claude Strava integration works through something called an MCP server — Model Context Protocol, the standard Anthropic uses to let Claude talk to external tools and data. Once connected, Claude reads your real Strava data: runs, rides, pace, heart rate, power, segments, everything. You get genuine AI analysis of your own training instead of generic advice.

There are two ways to set this up: the manual way, where you run your own Strava MCP server, and the shortcut, where you skip the technical setup entirely. This guide covers both honestly, so you can choose.

The manual way: run your own Strava MCP server

If you're comfortable in a terminal, you can set up the Strava AI integration yourself. The broad steps look like this:

You start by creating an API application in your Strava account settings, which gives you a Client ID and a Client Secret — the credentials the MCP server will use to read your data. Then you download and build an open-source Strava MCP server (there are a few on GitHub), which requires Node.js and a couple of build commands. Once compiled, you run the server locally and expose it so Claude can reach it, typically using a tunneling tool. Finally, you add the resulting URL to Claude as a connector.

It works, and it's free. But it's also a fair amount of moving parts: API credentials, Node.js, build steps, a server you have to keep running, and a tunnel URL that changes every time you restart it. For a developer doing it as a weekend project, it's an interesting setup. For everyone else, it's a lot of friction just to ask Claude about your last long run.

The shortcut (2 minutes, no terminal)

If you'd rather skip all of that, this is what we built AskYourData for. It runs the Strava MCP server for you — all you do is connect your accounts and paste one URL into Claude. Here's the whole flow.

Step 1 — Sign up and start your free trial

Registration is free, and so are your first 7 days. You add a card to start the trial, but you're not charged until day 7, and you can cancel anytime before then. ($2.99/month after that.)

Step 2 — Create your Strava API app

In your Strava account settings, you create a small API application — this is what lets the service read your Strava data securely on your behalf. You'll get a Client ID and a Client Secret, which you paste into your dashboard. Your data stays linked only to your account.

Step 3 — Get your setup email

Once that's done, you'll receive an email with your two personal links and simple instructions. Setup from here takes under 2 minutes:

First, connect your Strava account. The email contains an authorization link. You click it, you're taken to Strava, you click "Authorize", and that's it. You only do this once.

Then, add it to Claude. The email also contains your personal MCP URL. You add it to Claude as a connector:

⚠️ Keep your MCP URL private. It's linked to your Strava account and is yours only.

Step 4 — Start asking

Test it with something like "What are my recent Strava activities?" and Claude will pull them straight from your account.

One small tip: the first few times Claude uses the connector, it may ask you to confirm before reading your data. If your version of Claude offers an option to allow it to run without asking each time, turn that on — otherwise you'll be approving every single request, which gets old fast.

What you can actually ask Claude about your Strava data

Once the Strava AI integration is live, you can talk to your training data in plain English. A few examples people use regularly:

It works for any sport Strava tracks — running, cycling, hiking, triathlon, swimming, whatever you log. Claude can reason about any metric the Strava API exposes.

Manual or shortcut?

If you enjoy building things and want full control, the manual route is free and a genuinely interesting project. If you just want to analyze your Strava data with AI — without maintaining a server — the shortcut gets you there in a couple of minutes.

Frequently asked questions

What is a Strava MCP server for Claude?

A Strava MCP server is software that exposes your Strava data as tools Claude can call. It acts as a bridge between Claude and the Strava API, letting Claude read your activities, pace, heart rate, and other training metrics in real time during your conversations.

Is the Claude Strava integration free?

The manual route — setting up your own Strava MCP server — is free but requires technical setup. AskYourData's hosted version offers a 7-day free trial, then $2.99/month. You need a Claude account (free or paid) regardless of which route you choose.

What Strava data can Claude access?

Claude can access any data Strava makes available through its API: full activity history, pace, heart rate, power, elevation, distance, segments, and more. It works for all sports Strava tracks — running, cycling, hiking, triathlon, swimming.

Does the Strava AI integration with Claude work on mobile?

Yes. Add the connector on Claude Web (claude.ai) and it syncs automatically to the Claude mobile app on iOS and Android. No separate setup needed on mobile.