Installing feezify

← Back to docs index

Two ways in were planned. Only the developer path works today. I stopped developing feezify in September 2026.


A — No terminal (3 steps)

Works with Claude on the web and, best, Claude Cowork (desktop). Cowork can keep your data in a folder you own, so your journal and the copilot’s memory persist between conversations.

Requires a Strava account. The official connector is currently available to Strava subscribers. No Strava is fine — feezify still works from your journal alone (you just won’t get the objective training-load half).

In Claude: Customize → Connectors → add Strava → authorize (OAuth).

Step 2 — Install the skill

  1. Get feezify-skill-<version>.zip (never published: build it from a clone with pnpm package:skill, not tested on claude.ai, or use path B).
  2. In Claude: Settings → Capabilities → Skills → Upload skill.
  3. Make sure code execution is enabled (the skill runs a tiny bundled Node engine to compute your numbers deterministically).

The zip is self-contained — the engine, the core template, and the open method all ship inside it. Nothing else to install.

Step 3 — Ask for your read

Say: “how am I today?”

On the first run, the copilot creates your core (a folder of markdown you own — profile, journal, its own memory of you) and sets you up conversationally: your baselines, who you are, your consent for Strava, your first journal entry. It walks you through it one question at a time, like a friendly onboarding.

That’s it. Day to day: write a few honest lines in your journal, then ask for the read.

In Cowork, give Claude a folder (e.g. Documents/feezify) so your core persists. In a plain web chat without file access, the copilot will say so and offer a session-only read.


B — Developer path

Security rule (non-negotiable): adapter tokens live only in .env (git-ignored). Never write a real token into any other file. See AGENTS.md → Secrets and privacy & consent.

Prerequisites

Steps

# 1. Install & build
pnpm install && pnpm build

# 2. Create your portable core (your data, kept private)
cp -r core-template ~/my-feezify-core

# 3. Set up tokens (optional — only if you want Strava)
cp .env.example .env          # then add your Strava keys
git check-ignore .env         # must print ".env" — proves it's git-ignored

# 4. First read
node --env-file=.env dist/lecture.js ~/my-feezify-core 2026-06-29

Then fill your core:

When is Strava actually queried? Only when both the token is set in .env and connectors.strava: true in config.yml. Otherwise the engine reads an optional <core>/activities.json cache (shape: core-template/templates/activities-cache.md), or works from the journal alone.

Connect the skill to your AI

Point your AI at the right SKILL.md:

Same engine, different envelope.

Full checklist

The step-by-step, tick-as-you-go version lives in onboarding.md (🇫🇷 fr/onboarding.md). Onboarding is done when every box is checked.


Next → Daily use