FitOps CLI

Strava API & Data Use ​

FitOps reads your activity data from Strava through Strava's public API. This page explains how FitOps's design lines up with Strava's API Agreement and the November 2024 update — and why a personal, local-first tool sits squarely in what Strava explicitly continues to allow.

TL;DR ​

FitOps is a single-user, local-only tool. You connect with your own Strava API app under your own Strava account, and your data flows from Strava → your machine. Nothing is published, nothing is shared, nothing is used to train models, nothing is sold. The exact pattern Strava's announcement called out as remaining allowed: "coaching platforms focused on providing feedback to users and tools that help users understand their data and performance."

How FitOps Maps to the Agreement ​

Strava requirementHow FitOps complies
Display data only to the user it belongs to.FitOps runs entirely on your machine. Your activities are only ever shown to you, in your own terminal or your own browser at localhost:8888. There is no FitOps server, no shared feed, no public surface.
No use of Strava data for AI / ML model training.FitOps does not train any model. When you use FitOps with an AI assistant (Claude Code, Cursor, etc.), the assistant reads your data through the CLI to give you feedback — that's inference, not training. You choose the assistant; FitOps never ships your data anywhere on its own.
No public sharing or third-party redistribution of data.FitOps has no upload path, no telemetry, no analytics pipeline, no cloud sync (cloud backup, when enabled, writes to your Drive / OneDrive / GitHub repo — never to a FitOps service).
Cross-user / aggregate analytics are prohibited.FitOps is single-user by design. There is no cross-athlete aggregation anywhere in the codebase. The "analytics" in the name refers to your training metrics — CTL/ATL/TSB, VO2max, zones — computed from your activities, displayed to you.
Personal coaching insights are explicitly allowed.This is exactly what FitOps does: training-load tracking, zone analysis, weather-adjusted pace, race simulation — all classic personal coaching feedback, locally computed.
You authenticate with your own API credentials.The setup flow asks you to create your own Strava API application. You are the API-app owner; Strava's developer terms apply to you directly. FitOps is the local tool you use to work with the data Strava has agreed to send you.

What "Local-First" Really Means Here ​

Every byte of activity data lives in ~/.fitops/ on your machine:

~/.fitops/
├── config.json      # your Strava credentials (client id, refresh token)
├── fitops.db        # SQLite — all activities, streams, analytics
├── workouts/        # your workout markdown
└── notes/           # your training notes

There is no FitOps account. There are no FitOps servers. The project ships a CLI and a local web UI; both bind to localhost. If you delete ~/.fitops/, FitOps no longer has any of your data, instantly.

AI Agent Integration — Why It's Not "AI Training" ​

Strava's prohibition is on using Strava data to train artificial intelligence or machine learning models. FitOps's agent integration does the opposite:

  • The agent (Claude, Cursor, etc.) is a pre-trained model the user already chose.
  • The agent calls fitops CLI commands and receives JSON.
  • The agent reads that JSON to answer the user's questions about their own training.
  • Nothing is sent back upstream to train anything.

This is exactly the same shape as a coach reading a TrainingPeaks dashboard and giving you feedback — the model is the coach, your data is the input it considers in this one conversation, and Strava's announcement explicitly preserves this pattern.

What You Should Know ​

A few practical points so there are no surprises:

  • You bring your own Strava API app. Each FitOps user creates one through strava.com/settings/api. Strava's "Single Player Mode" defaults to a 1-athlete cap, which is exactly what FitOps needs.
  • Rate limits are Strava's, not ours. 200 requests / 15 min and 2,000 / day. The sync engine batches and respects these.
  • You control retention. Activities are stored locally for as long as you keep them. Removing the local database removes all of it.
  • Your LLM choice is yours. If you point an agent at FitOps, the data leaving your machine is whatever you choose to put in your prompts — a decision that lives between you and your LLM provider, not FitOps.

References ​

← Back to Concepts

FitOps CLI — terminal-first training analytics