COROS MCP: Connect your training data with ChatGPT and Claude

On May 19, 2026, COROS unveiled a genuine first: a secure, standardised and privacy-respecting connection between your training data and the two most widely used AI assistants, ChatGPT and Claude. It's a bold move that puts COROS among the pioneers of this approach. Below we'll walk through the details - what this integration actually gives you, and where its current limits lie.

Introduction

The latest surveys show that the vast majority of people now use AI tools - for everything from quick everyday questions and technical troubleshooting to getting their head around complex topics and making real, practical decisions. COROS is the first brand in the outdoor watch segment to build a secure, standardised and privacy-respecting way to open up your training data to the two most popular AI assistants: ChatGPT from OpenAI and Claude from Anthropic.

How it worked until now

Until now, the routine usually went like this: you'd take a screenshot of a workout, or export your training data in a format like .fit, and hand it over to something like ChatGPT to interpret. The drawback is obvious - the whole process is manual, and access to your historical data is patchy and inconsistent.

And yet context is one of the most important ingredients in sport. The data from a single workout isn't an isolated island. It lives within the context of your athletic life, your short- and long-term goals, the stress you carry in everyday life, the quality of your sleep, and other objective physiological markers such as HRV (heart rate variability), resting heart rate, the state of your parasympathetic nervous system, and so on.

How it works now - the MCP protocol

Connecting AI assistants (built on so-called LLMs - Large Language Models) to large datasets has been common practice for years in plenty of other fields. Medicine and finance, for example, where AI helps surface trends, interpret them and uncover new connections.

The modern, standardised way to give AI access to that data is a protocol called MCP (Model Context Protocol). It lets AI reach third-party information securely, under the following conditions:

• It's a secure connection governed by standardised rules.
• Your privacy is preserved - you and only you grant access to the data, explicitly and knowingly.
• Those permissions can be managed and revoked at any time.

And that's exactly the kind of MCP access COROS has offered since May 2026. It's an unprecedented step toward democratising and diversifying how you work with your own data and interpret it - well beyond what the COROS app alone provides.

COROS is the first sports watch brand in the world to officially support MCP. The connection opens up 15 so-called endpoints (access points) across five categories: profile and devices, activities, daily health metrics, EvoLab assessments (HRV, recovery, training load, VO₂ Max, Race Predictions) and your training plan.

What this means for you

The data your COROS watch collects is interpreted through the COROS EvoLab toolset, which places it in a wider context and respects the broadly accepted training and physiological principles of how the human body adapts under load. But COROS has chosen to leave the final call to you - how you go on to work with that data and interpret it.

To put it plainly: we're all a little different, and each of us prefers a different way of making sense of information. Some people like clean, colourful diagrams, rings and charts. Others prefer to read the story in the hard numbers. And others want a written interpretation that explains the measured trends in a way that simply clicks.

Tools like ChatGPT and Claude give you that flexibility, because the way they communicate adapts to the style you prefer. If you'd rather see your training data visualised across three charts - that works. If you'd rather have a written summary in a single paragraph, ChatGPT or Claude handles that just as easily.

On top of that, your question about progress can be far more nuanced and unfold as an ordinary conversation, in whatever way suits you. And now that ChatGPT and Claude have secure access to all of your data - including the past - the interpretation is far more holistic and joined-up.

The conversation doesn't have to revolve only around interpreting the past, either. Your questions can point forward, too - how to adjust your training, what to steer clear of, how to time your peak ahead of an upcoming race, and so on. The only limit on what you can do with the data is your own curiosity and imagination.

And finally - ChatGPT and Claude are currently among the most advanced AI tools out there, capable of working across the enormous body of knowledge humanity has gathered and synthesising it in unexpected, original and inventive ways.

Another way to look at it: the data your COROS watch collects is only the first step. Interpreting it meaningfully is the far more important one. Opening that data up to the two most capable AI tools available takes the interpretation to a whole new level.

What you need to get started

At the time of writing, the data is available through the paid ChatGPT Plus (from OpenAI) and Claude Pro (from Anthropic) plans, since those are currently the only tiers that allow an MCP connection. Both cost around 20 euros a month.

If you're more technically inclined, you can also set up the connection through the developer tool Cursor, as well as through Google Gemini - though Gemini only via its CLI (command line) interface.

The MCP connection is available to every user with a COROS watch and an active COROS account - whether that's the entry-level PACE 4 or the flagship APEX 4.

How to set up the connection

The first step is to get the right MCP link for your region from the official COROS site:

• Europe: https://mcpeu.coros.com/mcp

ChatGPT

1. Open ChatGPT and go to Settings – Apps – Advanced Settings.
2. Turn on Developer Mode.
3. Click Add App.
4. Paste the MCP link for your region.
5. Authorise and sign in to your COROS account.
6. Done - start asking.

Claude (desktop app)

1. Open Claude and click the gear icon (Settings).
2. Go to Connectors – Add Custom Connector.
3. Paste the MCP link for your region.
4. Sign in to your COROS account when prompted.
5. Set the access permissions so your data can flow through the connector.
6. Click Save.
7. Done - start asking.

Frequently asked questions

I'm getting "connection failed". What now?

Check that you've used the correct MCP link for your region (see the list above).

It asks me to sign in, but nothing happens.

Try signing in directly at coros.com first, to make sure your COROS account is working.

I don't see "MCP Servers" / "Connectors" in the settings.

To unlock the MCP feature you most likely need a paid plan (ChatGPT Plus or Claude Pro). The free versions usually don't support it.

Which platforms currently support MCP?

ChatGPT, Claude and Cursor have full support. The process is very similar across all of them - look in the settings for something like Apps, Connectors or Extensions.

What about Google Gemini?

The regular Gemini chat app (gemini.google.com) doesn't currently support custom MCP connectors the way ChatGPT and Claude do. If you're comfortable working in a terminal, though, you can set up the connection through the Gemini CLI - you'll find the steps in Google's documentation, and you use the MCP link for your region as the server address. If Google adds MCP support to the regular Gemini app in the future, we'll update this guide.

Privacy and security

This connection creates no new security risks beyond those that already exist when you use your COROS account and your chosen AI platform on their own. It's a structured layer that lets two systems talk to each other in a controlled way, based on permissions you grant explicitly.

In practical terms, that means:

• Your COROS data is still protected by COROS security standards. Your interactions with the AI are still governed by ChatGPT's or Claude's rules.
• The MCP connection doesn't bypass any authentication or existing privacy controls. There's no "back door" here.
• Access is granted solely on your explicit authorisation, and only to the extent you agree to. You can revoke it at any time.
• At launch the integration is read-only - the AI can't write or change anything in your account.

One interesting detail - COROS has chosen not to expose the most granular data through MCP. The AI doesn't receive detailed GPS tracks, elevation profiles, second-by-second heart rate and pace records, power meter data, or lap and interval details. It works with summary data and trends - which is more than enough for meaningful interpretation, while keeping an extra layer of privacy intact.

In short - think of this connection as a secure extension of the tools you already use. Your data protection stays fully in force, your permissions still call the shots, and you keep complete control over what's connected and what the AI is allowed to do.

What's coming next

At launch, the MCP integration is limited to reading data. But COROS has already announced that write permissions are coming in the next phase. That will mean AI tools can, for example, generate a training plan and write it straight into your calendar, or adjust planned workouts based on your current recovery and upcoming races. In other words - a truly personal AI coach that doesn't just advise, but can act.

Conclusion

The step COROS has taken is one of the most interesting we've seen in a while. The watch and the app remain what they always were - a reliable way to collect and interpret your data. But as of May 2026, there's now the added option to take that data and open it up to the most advanced AI tools around. Securely, transparently, and fully under your control.

Instead of exporting .fit files or screenshotting charts, you can hold an ordinary conversation about your training with an AI - and the answers come back in the context of everything your data knows about you. It's worth a try.

If you bought your COROS watch from the ralpu store and this whole connection feels complicated or you're not sure where to start, don't hesitate to get in touch. Drop us a message via the contact page and we'll walk through it together, step by step, so these new capabilities are within your reach too.