How I built my own Personal AI Coach — For Free and stopped paying for fitness apps

hari balaji — Tue, 02 Jun 2026 17:02:44 +0000

A bit about me

I am a regular runner and I use these apps frequently - Strava, Runna, and Garmin connect.

These are the gears I use:

Garmin vivoactive 4
Skechers go run consistent

My workflow

I record my activity using my watch and it uploads it to Garmin connect. I've already enabled the Garmin-Strava integration so all activities are automatically pushed to Strava. For workout plans, I used to rely on Garmin Connect's workouts and Runna's workouts. The former is free but the experience is not very intuitive. And Runna's was very good and is very interactive. But you have to pay Runna to access it. While I enjoyed my trial period with Runna, I did not want to pay for both Strava and Runna.

And I ended up setting up an Agent that acts as a Coach who's available 24x7 on telegram.

This is how I have been using it:

I tell my goal and get a workout plan. This is calculated based on my previous activities

Last week I had to travel all of a sudden which disrupted my workout plan. So asked it for a modified plan.

I ask it to review my latest activity

I spoke to it literally. I informed that my legs are sore and asked it's advice. This is what it had to say to me:

This suits very well for me and now, onto the setup.

How I Set Up My AI Coach

Step 1: Installed Picoclaw

Picoclaw is the brain of the operation. It's a lightweight, open-source AI assistant that runs locally on your machine. I followed the installation guide to get it running on my setup.

Picoclaw connects to your AI model of choice and gives you a local chat interface — think of it as your coach's "home base."

Step 2: Added Skills

Skills extend what the coach can do. Here's what I installed:

Garmin Health Analysis — [Install]

I use a Garmin watch, so this skill was essential. It connects my Garmin data to the AI coach. Once installed, my coach could:

Read my workout history — every run, ride, swim, and walk
Analyze metrics — heart rate zones, pace trends, sleep quality, Body Battery, HRV, and more
Track progress over time — "How's my 5K pace trending over the last month?"

My coach uses this data to make informed decisions about my training plan. It's not guessing — it's looking at my actual numbers.

Weather — [Install]

I added this one so the coach would know the weather before building my daily plan. Now it factors in temperature, humidity, and conditions — so my interval session doesn't land on the hottest day of the week.

macOS Local Voice (optional, recommended if you use a mac) — [Install]

Typing after a run with sweaty fingers is annoying. This skill lets me send voice messages that the coach transcribes and responds to — using Apple's native Speech framework, fully offline. No API keys needed.

Step 3: Connected a Telegram Bot using the below steps

Open Telegram and search for @botfather
Send /newbot and follow the prompts to create a bot
Copy the API token BotFather gives you
Add the token to your Picoclaw config
Restart Picoclaw

And that's it !

The whole setup takes about 30 minutes. You get a coach that knows your body, checks the weather, adapts to your life, and never charges you a rupee.

DEV Community: hari balaji