You track your meals in one app. Exercise in another. Sleep in a third. Weight in a spreadsheet. Let's fix that.

There are hundreds of health tracking apps. MyFitnessPal. Fitbit. Sleep Cycle. Lose It. Noom. MyFitnessPal alone has 200 million users.

And yet we still can't answer simple questions like "Do I sleep better on days I exercise?"

Why? Because these apps don't talk to each other.

What if we didn't need four separate apps? What if we taught our AI to track our health instead, all in one place, all connected?

That's what we're building today. A personal AI with four health skills: meals, exercise, sleep, and weight. When we're done, you can delete the apps.

What Health Apps Got Wrong

We analyzed reviews across all the major health apps. Here's what people complain about:

Everything is siloed. Your meals app doesn't know you exercised. Your exercise app doesn't know you slept terribly. You're the integration layer—in your head.

Subscription overload. MyFitnessPal Premium: $80/year. Fitbit Premium: $80/year. Sleep Cycle Premium: $40/year. That's $200/year minimum, and they still don't connect.

Your data sits on their servers. MyFitnessPal has 200 million users' food logs in their database. That's their business model. Sure, you can export to CSV—but then you lose all the app logic.

Manual entry is tedious. Log every meal. Start every workout. Remember to track sleep. Miss a few days and you give up.

No cross-domain insights. "Do I lose more weight when I sleep 8 hours?" "What's my net calories this week?" These questions require multiple apps and a spreadsheet.

What if your AI could see the whole picture because all your health data lived in one place?

Skills Instead of Apps

Here's the shift. Instead of downloading health apps, we teach our AI health skills.

Old way: 4 apps on your phone, $200/year in subscriptions, data scattered across company servers.

New way: One AI that you teach. Four connected skills. Your data in your vault. Portable. Private. Yours.

What Health Skills Look Like

A personal AI with health skills has two parts for each skill:

1. The Data Tables

This is where your actual data lives. Four simple tables:

Meals Table

Column

What it stores

id

Unique identifier

food

What you ate

calories

Calorie count

protein_g

Grams of protein

fat_g

Grams of fat

carbs_g

Grams of carbs

meal_type

Breakfast, lunch, dinner, snack

created_at

When you logged it

Exercises Table

Column

What it stores

id

Unique identifier

exercise

Type of workout

duration_minutes

How long

calories_burned

Calories burned

notes

Optional notes

created_at

When you logged it

Sleep Table

Column

What it stores

id

Unique identifier

sleep_date

The night of

bedtime

When you went to bed

wake_time

When you woke up

total_hours

Total sleep

quality

Rating 1-5

created_at

When you logged it

Weight Table

Column

What it stores

id

Unique identifier

weight_lbs

Your weight

notes

Optional notes

created_at

When you logged it

These get stored as Parquet files in your private Arca vault. Not on MyFitnessPal's server. Not mixed with 200 million other users. You own it. Export it anytime.

2. The SKILL.md Files

Each skill gets an instruction manual for your AI. It tells your AI:

  • What is this data?

  • What are some example queries?

  • How do these skills relate to each other?

The magic: your AI can query across all four tables. "What's my net calories?" pulls from meals AND exercises. "Do I sleep better when I work out?" correlates sleep AND exercise data. No app can do this.

Let's Build It

Here's how to build your personal AI with health tracking skills using n8n and Arca.

What You'll Need

  • An n8n account (cloud or self-hosted)

  • An Anthropic API key (for Claude)

  • An Arca account (Get free at arca.build)

Step 1: Import the Template

I have a template workflow ready. In n8n start a new workflow, then:

  1. Go to Import

  2. Select Import from URL

  3. Paste this URL: https://arca.build/health_n8n_template.json

  4. The workflow will load with all the components

Step 2: Connect Your Claude Account

  1. Double-click on the Claude node

  2. Click Create New Credential

  3. Go to console.anthropic.com

  4. Create an API key

  5. Paste it into n8n

  6. Save

We use Claude because it's great at MCP tool calling.

Step 3: Create Your Arca Account

  1. Go to arca.build

  2. Sign up with Google or email

  3. You'll land on an empty dashboard—that's fine

  4. Go to your Account page

  5. Copy your API Key

Step 4: Connect Arca to n8n

  1. Find the MCP Server node in your workflow

  2. The endpoint should be: https://mcp.arca.build

  3. Authentication: Bearer Auth

  4. Click Create New Credential

  5. Name it something like "Arca API Key"

  6. Paste your Arca API key from Step 3 as the Bearer token

  7. Save

Step 5: Select the MCP Tools

In the MCP Server node, you'll see available tools:

Select all of them and save.

Step 6: Activate and Test

  1. Turn on the workflow (toggle in top right)

  2. Copy the webhook URL that appears

  3. Open it in a new browser tab

This is your personal AI interface. Bookmark this on your phone's home screen.

Step 7: Create Your Health Skills

In the chat interface, create each skill through conversation:

Create the meals skill:

I want to track my meals with calories and macros. 
Create a meals skill with food name, calories, protein, fat, carbs, and meal type.

Log your first meal:

I had a chicken salad for lunch - about 450 calories, 35g protein, 15g fat, 20g carbs

Create the exercises skill:

Now let's track exercise. Create an exercises skill with exercise type, 
duration in minutes, and calories burned.

Log a workout:

I did a 30 minute run and burned about 300 calories

Here's the magic moment. Ask:

What's my net calories today?

Your AI will query both tables and calculate: 450 calories consumed - 300 burned = 150 net calories.

No combination of separate apps can do this.

Create the sleep skill:

I want to track sleep. Create a sleep skill with sleep date, bedtime, 
wake time, total hours, and quality rating 1-5.

Log your sleep:

Last night I went to bed at 11pm, woke up at 7am, got about 7.5 hours, quality was a 4

Create the weight skill:

Finally, let's track weight. Create a weight skill that stores weight in pounds.

Log your weight:

I weighed 185 this morning

Step 8: Verify in Arca

Go back to your Arca dashboard:

  1. Click Home

  2. Check Data Tables, you'll see meals, exercises, sleep, weight

  3. Click View Entries on any table to see your data

  4. Check Skills, you'll see the SKILL.md files auto-generated for each

Using Your Personal AI

Now you can interact naturally:

Meal tracking:

  • "I had two eggs and toast for breakfast, maybe 350 calories"

  • "What have I eaten today?"

  • "How many calories this week?"

Exercise tracking:

  • "I did 45 minutes of cycling, burned about 400 calories"

  • "What workouts have I done this week?"

  • "How many calories have I burned today?"

The connected queries (the magic):

  • "What's my net calories today?"

  • "What's my average net calories this week?"

  • "On days I exercise, do I sleep better?"

  • "Give me a health summary for this week"

Sleep tracking:

  • "I slept from 10:30pm to 6am, quality 3"

  • "How's my sleep been this week?"

  • "What's my average sleep quality?"

Weight tracking:

  • "I weighed 183 this morning"

  • "What's my weight trend this month?"

  • "Am I making progress?"

Your AI reads the skills, queries your data across all four tables, and responds. No app to open. No UI to learn. One conversation.

Connect Everywhere

Your Arca vault works with any AI that supports MCP:

  • Claude - Add Arca as a connector in Claude's settings

  • ChatGPT - Use the custom connector feature

  • Your n8n workflow - What we just built

  • Any MCP-compatible agent - Same API key, same data

Your data follows you. Switch AI providers whenever you want. Your health history comes with you.

Check https://arca.build/mcp to learn how to connect.

Don't Want to Build It?

That's fine. If you'd rather skip the setup and just use a personal AI that already has all of this built in, check out Mio.

Mio is the home of your hosted personal AI that uses Arca as its data layer. You sign up, give your AI a name, and start teaching it skills through conversation. No n8n. No API keys. No configuration.

You can talk to Mio through SMS, WhatsApp, phone calls, or just the website. And if you want to use it inside Claude or ChatGPT, Mio has its own MCP server you can connect.

Same idea. Same data ownership. Someone else did the building.

What's Next

This is just the beginning. In upcoming episodes, we'll add more skills:

  • Journal entries and mood tracking

  • Recipe storage and meal planning

  • Grocery lists that learn what you buy

  • Habit tracking

  • And more

Each skill builds on the same pattern: a data table and a SKILL.md file in your Arca vault. And because they're all in the same vault, they can talk to each other.

The goal is to replace the apps on your phone with skills in your personal AI. One AI instead of 100 apps.

Questions?

Drop a comment on the video or DM me on LinkedIn. Let's build this together.

Keep Reading

No posts found