Analyse Your Garmin Data With AI — Without Setting Up MCP
The MCP route gives Claude live access to your Garmin account, but it takes a terminal and a bit of patience. Here's the simpler way — export your data once, point Claude at the folder, and run one of two ready-made prompts: a 90+ sleep score protocol, or a VO2max plateau diagnosis.
Analyse your Garmin data with AI, without the MCP setup
I've written before about connecting Garmin to Claude through MCP, which gives the AI live access to your account. It's the better long-term setup — but it involves a terminal, and for a lot of people that's where it ends.
Here's the simpler route: export your data once, point Claude at the folder, and run a prompt. No live connection, no tokens, and it works with the desktop app you may already have.
Step 1: Export your Garmin data
Go to garmin.com → Account → Data Management → Export Data.
- Log in to your Garmin account
- Click "Export Your Data"
- Garmin will prepare a ZIP file — this can take a few minutes to a couple of hours. You'll get an email when it's ready.
- Download and unzip it. Look for the DI_CONNECT folder — that's where your sleep, HRV, stress, body battery, resting heart rate and activity data all live.
Step 2: Choose your tool
You need a version of Claude that can read files directly from your computer — not the regular chat at claude.ai. You have two options:
Option A: Claude Cowork (easier — no terminal needed)
Claude Cowork is a desktop app for Mac and Windows. It works like Claude Code but without any command line — you just point it at a folder and describe your task.
- Download and install the Claude desktop app from claude.ai/download
- Open it and log in with your Claude account
- Grant Cowork access to your unzipped Garmin export folder when prompted
- Paste the prompt from Step 3 and let it run
Requires a paid Claude plan (Pro at $20/month).
Option B: Claude Code (more powerful — requires a terminal)
Claude Code is the developer-focused version. It's slightly more setup but gives Claude full access to the folder structure.
Open your terminal
On Mac: Press Cmd + Space, type Terminal, and press Enter.
On Windows: Press the Windows key, type PowerShell, right-click it and choose "Run as administrator".
Install Claude Code
Paste this into your terminal and press Enter:
curl -fsSL https://claude.ai/install.sh | bashText will scroll by — that's normal. When it stops and shows a new cursor, it's done.
Log in and navigate to your folder
Type cd (with a space), then drag the unzipped Garmin folder from Finder or Explorer into the terminal window. Press Enter. Then start Claude Code:
claudeIt will open a browser window to log in — use your existing Claude account.
Step 3: Pick a prompt
This is where it lives or dies. Ask "how do I sleep better" and you'll get the same eight bullet points everyone gets. Give it a structured brief and it works from your actual numbers. Here are the two I use most.
Prompt 1 — Hit a 90+ sleep score
Runs a deep analysis of your sleep data and returns a structured report with a 30-day protocol, built around the nights you already score well on.
You are an expert sleep coach and sports scientist with deep knowledge of Garmin's sleep scoring algorithm and sleep physiology. I want to consistently achieve Garmin sleep scores of 90 or above. I'm giving you my full Garmin data export. Please read all relevant JSON files you can find — sleep, HRV status, stress, activities, resting heart rate, body battery — whatever is available. Most interesting files are in the DI_CONNECT folder. --- Perform a DEEP PERSONAL ANALYSIS and produce a structured report with these sections: ### 1. Current sleep profile My current sleep quality baseline — average score, consistency, trends over time. Be specific with numbers from my data. ### 2. What happened on my 90+ nights Analyze every night where I scored 90 or above. What patterns do you see? Compare sleep duration, HRV, sleep stage distribution (deep %, REM %, light %), bedtime, wake time, and stress vs. my average nights. What did I do RIGHT on those nights that I didn't do on lower-scoring nights? ### 3. Key factors holding me back Based specifically on my data, what are the top 3–5 barriers preventing consistent 90+ scores? Look at: sleep duration, bedtime/wake time consistency, HRV trends, stress levels, activity timing, deep sleep %, REM %, and any red flags. Be direct and reference my actual numbers. ### 4. Personalized action plan (prioritized by impact) A concrete, ordered action plan — most impactful first. For each action: - What exactly to do (specific and measurable) - Why it matters for MY score specifically (reference my data) - Expected timeline to see results ### 5. Weekly routine template A Mon–Sun routine optimized for 90+ sleep scores, based on my actual activity patterns and lifestyle signals visible in the data. ### 6. HRV & recovery rules What my HRV trend means for my sleep quality. Give me specific HRV-based decision rules (e.g. "On days your HRV is below X, do Y instead of Z"). ### 7. 30-day protocol A specific 30-day plan to break through to consistent 90+ scores, with week-by-week milestones based on where I currently am. Be direct, data-driven, and specific to my actual patterns. Don't give me generic sleep hygiene advice unless my data shows it's actually relevant to my situation. Reference research where useful but keep everything practical and actionable.
Prompt 2 — Work out why your VO2max is stuck
If you train consistently but your VO2max won't move, this one works through your intensity distribution, recovery and training load in three phases to find the actual cause.
You are a data-driven endurance coach and sports scientist. I train multiple times a week, but my VO2max has plateaued — it is not improving. I have attached my full Garmin data export. The most important data is in the DI_CONNECT folder. It contains: activities (runs, rides, walks, etc.), sleep data, HRV status, body battery, stress scores, training load, training readiness, VO2max history, intensity minutes, and recovery metrics. Your job is to figure out WHY my VO2max is stuck and what to change. Work through three phases. Complete each fully before moving to the next. PHASE 1 — ANALYSE Read all files inside the DI_CONNECT folder. Then extract: My VO2max history and how long it has been flat or declining My average sleep quality, duration, and HRV trend My weekly training load: volume and intensity distribution My training consistency — gaps, monotony, overtraining signals, recovery patterns My intensity balance: what percentage of my activities fall in Zone 1–2 (easy) vs Zone 3 (grey zone) vs Zone 4–5 (hard)? Any red flags: chronic fatigue, high resting HR, suppressed HRV, under-recovery Present this as a concise athlete profile with specific numbers. PHASE 2 — DIAGNOSE THE PLATEAU Based only on my data, identify the specific reasons my VO2max has stalled, ranked by impact. Consider the usual causes and tell me which actually apply to ME, citing my numbers: Too much moderate "grey zone" training and not enough true easy Zone 2 volume Not enough genuine high-intensity VO2max work — or too much of it without recovery No progressive overload: the same sessions and the same load week after week Poor sleep, suppressed HRV, or chronic under-recovery blunting adaptation Training monotony or accumulated fatigue masking my real fitness For each cause that applies, point to the data that proves it. PHASE 3 — HOW TO BREAK THE PLATEAU Give me a prioritised set of changes to restart progress, based only on what you found: The single biggest lever I should pull first Concrete adjustments to my weekly structure: intensity distribution, key workouts, recovery Sleep and recovery targets based on my actual numbers What to stop doing, not just what to add End with: the one change most likely to unstick my VO2max, and the single metric to watch each week to confirm it is working.
Why this beats asking a chatbot cold
Ask any AI a generic health question and you get generic advice: dark room, no screens, consistent bedtime. Useless, because it isn't about you.
Point it at your own export and it's working from your bedtimes, your HRV, your training load, your resting heart rate — measured on your wrist every night for however many years you've owned the watch. It can tell you which of those generic tips your data suggests would actually move the needle, and which you're already doing fine on.
I used exactly this on my own sleep, got specific and unglamorous guidance out of it, and it worked.
Two caveats
It's a snapshot, not a connection. The export is frozen at the moment you downloaded it. Ask again in two months and you're analysing stale data — you'd need to export again. That's the one real advantage of the MCP setup: it always sees today's numbers.
It's your full health history. Treat the folder accordingly, and be deliberate about what you upload and where.
If you'd rather have the live connection and don't mind ten minutes of copy-paste in a terminal, the MCP route is written up here:
Join the Garmin School
Leave your email to get access to the Garmin School and receive updates on new videos and apps.
What to read next
Day 12846How to Track a Hyrox Properly on Your Garmin
Eight runs and eight stations is sixteen legs — but you don't build a sixteen-step workout. You build two activities and let them loop, and you get a clean split for every run and every station.
Day 12720Put Your parkrun Barcode on Your Garmin and Never Forget It Again
No barcode, no time — that's the parkrun rule. If you hate running with your phone, you can store the barcode on your watch instead, and the same trick works for gym and loyalty cards.
