MCP

The Best Claude & ChatGPT Connectors for Cyclists and Endurance Athletes (2026)

Joules June 26, 2026 · 8 min read

In the last few months a real category appeared: connectors that plug your training data and tools straight into the AI assistant you already pay for. Add one once and Claude or ChatGPT can work with your actual rides instead of vague advice. Here's an honest field guide to the ones worth adding if you train seriously — sorted by the job they actually do, with the trade-offs, not a hype list.

First, what a "connector" is — and the one gate

A connector (built on MCP, the Model Context Protocol) is a standard way to give an AI assistant live access to a data source or tool. Instead of exporting CSVs and pasting them into the chat, you authorize the connector once and the assistant can query it directly inside a normal conversation. Worth stating up front: on Claude, custom connectors work on every plan — including the free tier (free Claude allows one custom connector). On ChatGPT, they need a paid plan (Plus/Pro).

The useful way to sort the field is by which side of training a connector lives on. Training has a read side (what did I do?) and a write side (what should I do next, and did it work?). Most connectors today live on the read side, and a handful of coaching connectors are starting to fill in the write side. Here's the map.

The read side — make the AI an analyst of your history

Strava's official MCP connector

Strava shipped its own MCP connector in June 2026, available to subscribers. Authorize it from Claude's connector settings and the assistant can read your activities — heart rate, power, pace, GPS — and reason over them in plain language: summarize a block, spot a trend, pull up your best 20-minute power this month. It's the cleanest read side going because it comes straight from the source, with Strava's own guardrails.

• Best for: anyone who already lives in Strava and wants a smarter lens on what they've already done.
• The limit: it's read-only by design. It analyzes the past; it doesn't prescribe tomorrow or build a plan. That's the line between analysis and coaching, and the connector sits firmly on the analysis side.

We wrote a deeper breakdown of exactly where that line falls in Strava's MCP Connector for Claude: What It Does — and What It Can't (Yet).

Multi-source aggregators (e.g. athletedata.health)

If your data is scattered across Strava, Garmin, WHOOP, Oura, Hevy, Withings and Apple Health, aggregator connectors pull many sources behind a single connection and expose them all as tools the AI can query live. The pitch is breadth: one authorization, twenty-plus sources, no manual exports.

• Best for: data-rich athletes who want recovery, sleep and training in one place and like to ask cross-source questions ("did my HRV dip the week my volume spiked?").
• The limit: still fundamentally the read side. Aggregation makes the analysis richer and wider, but the assistant is describing your data, not committing you to a plan and holding you to it. More inputs, same job.

The write side — a coach that plans your future

Reading history is the easy half to ship, which is why most connectors start there. The harder half is the write side: an assistant that interviews you to build an athlete profile, sets a season backward from a goal race, prescribes each week with intent, and grades what you actually did so next week adjusts. That's coaching — a different category from analysis — and a few coaching connectors are now emerging to do it. What separates them is friction: whether a coach makes you bring a separate platform, subscription, or device, or whether it coaches from a short conversation inside the assistant you already pay for.

That write side is what Joules is built for. It's an MCP app that gives Claude (or ChatGPT) a cycling coach's brain: a short interview, a season built around your A-race, structured weeks, and a log so the plan progresses and stays honest — device-free, just duration and perceived effort if that's all you've got. There's no separate app to download and no device you're required to own: you add it the same way you add Strava, by pasting the connector URL once, and it coaches inside the Claude or ChatGPT you already pay for.

• Best for: riders who don't just want to understand their training — they want a plan for the next twelve weeks and a coach that adjusts it when life interrupts.
• The trade: it asks you a few questions up front (a coach has to know your goal and constraints), and full access is free while Joules is in beta. In exchange you get prescription and accountability, not just a readout.

How to choose — and why you don't have to

The honest framing isn't "which connector wins." It's "which job do you need," and the best setups stack a read connector with the write one in the same chat:

• Just want a smarter view of your history? Add Strava's connector (or an aggregator if your data is spread out). Done.
• Want the AI to actually tell you what to ride tomorrow, toward a goal? Add a coaching connector like Joules — that's the side history can't cover.
• Want both? Run them together. Strava tells the assistant what you did; Joules tells it what to do next, then checks whether you did it. Read side and write side, one conversation.

The bottom line

The connector wave is real and worth riding: the AI you already pay for can now work from your actual training instead of generic advice. Just know what each one does. The read side — Strava's MCP, the aggregators — turns the assistant into a sharp analyst of your past, and that's genuinely useful. The write side turns it into a coach that plans forward and holds you to it. They're not competitors; they're two halves of the same loop, and you can run both at once.

New to all this? Start with Can AI Be Your Cycling Coach? A Practical Guide for 2026, or see a real Joules training week — day by day, no install required.