SECTION 11

Getting started

From the maintained repository documentation.

On this page

Quick Start

Recommended: Paste SETUP_ASSISTANT.md into ChatGPT, Claude, Gemini, or any AI chat, and it will walk you through everything below step by step. This is the easiest path for most users.

You can also follow the step-by-step guides below.

1. Create Your Dossier: your stable private context

Copy DOSSIER_TEMPLATE.md and fill in the context your AI cannot read from your data: background, equipment, stable constraints, health and injury history, tested fueling, and how you prefer to be coached. Current thresholds and load always come from your JSON. You can start coaching before the dossier exists; a missing dossier limits personalisation, not safety.

2. Set Up Your Data Sync

Keep your Intervals.icu data fresh for your AI coach automatically.

Local sync: a script on a machine you control syncs your data on a 60-second timer. An AI whose runtime can reach that filesystem reads the files directly; otherwise it reads them through a cloud connector (Google Drive, OneDrive; platform support varies).

GitHub sync: GitHub Actions syncs to a private repo every 30 minutes by default, and Sync Now requests a sync when you want one, for example after a workout; it waits if another sync is already running or queued. Scheduled runs are best-effort, and each run uses your GitHub Actions minutes; see Sync Schedule and GitHub Actions Usage for Fast mode and hybrid schedules. A hybrid schedule checks more often during an active window fixed in UTC, so its local-clock hours can shift when daylight saving time starts or ends. Your AI reads the data through a GitHub connector; raw URLs work only for a public repository.

On-demand sync: trigger a fresh sync from your phone or browser via your repo's README. Download the data as a ZIP artifact. No schedule, no local Python.

Manual export: run once, upload the file. No automation. Custom ranges available. Remember the upload is frozen. Re-export and replace it before a report if your data has moved on.

3. Configure Your AI Platform

Choose your path:

  • Agentic Platforms: OpenClaw, Claude Code, Claude Cowork, ChatGPT Codex, Gemini CLI (code execution, push workouts to calendar). Grok Bot and Hermes Agent have the capability class but are experimental; capability class is not a support promise. Calendar writes need verified access, configured credentials and a tested integration.
  • Web Chat Platforms: ChatGPT Projects, Claude Projects, Gemini Gems, Grok (web/app), Mistral Vibe

4. Make Files Available to Your AI

Your AI needs access to SECTION_11.md (the protocol), plus DOSSIER.md (your stable private context and preferences) if you use one. How depends on your setup:

  • Local/agentic: Where the runtime can reach your filesystem, the AI reads the files directly.
  • GitHub connector: If these files are in your connected repo, the AI reads them directly. If only DOSSIER.md is in your data repo, upload SECTION_11.md separately (or connect the CrankAddict/section-11 repo too).
  • Cloud connector (Google Drive, OneDrive; platform support varies): If these files are in your synced folder, the AI reads them through the connector.
  • URL fetch (no connector): If your data repo is public, the AI can fetch raw URLs directly.
  • Manual upload: Upload SECTION_11.md, and DOSSIER.md if you use one, to your AI platform or project. Uploaded files are frozen at upload. Replace the old copy when you update one, and don't leave two versions in the store.