One framework.
Every fitness format.
A TypeScript framework for converting between FIT, TCX, ZWO, and Garmin Connect formats. Use the visual editor or build with the SDK.
npm i @kaiord/core
Create and manage workouts
A complete toolkit for athletes, coaches, and fitness enthusiasts.
Visual Workout Editor
Drag-and-drop interface to create structured workouts with steps, intervals, and targets. Runs entirely in your browser.
AI Workout Generation
Describe your training goal in plain language and generate structured workouts with Claude, GPT, or Gemini.
Garmin Connect Sync
Push workouts directly to Garmin Connect from your browser. No server needed — works via a lightweight browser extension.
One hub. Every format.
KRD is the canonical format. All conversions flow through it — lossless, round-trip safe.
Built for developers
Convert fitness data in 4 lines of TypeScript. Strategy pattern — bring your own adapters.
import { fromBinary, fitReader } from '@kaiord/fit'
import { toText, tcxWriter } from '@kaiord/tcx'
const krd = await fromBinary(fitFile, fitReader)
const tcx = await toText(krd, tcxWriter)TypeScript-first
Strict types, Zod schemas, full IntelliSense. No
any.
Hexagonal Architecture
Clean boundaries. Domain never depends on infrastructure.
Plugin System
Strategy pattern: inject your own readers and writers.
CLI
Convert files from your terminal. Pipe-friendly.
MCP Server
AI-ready via Model Context Protocol. Works with Claude, Cursor, and more.
5 Format Adapters
FIT, TCX, ZWO, GCN, KRD. All round-trip safe.
Every line written by AI agents
This entire project — domain logic, adapters, CLI, MCP server, web editor, tests, CI/CD, and this landing page — was written by Claude Code. Not a single line of human-written code. A real-world showcase of AI-assisted development at scale.
See the commit history →No servers. No accounts. No cloud.
Everything runs locally or in-browser. The CLI converts files on your machine. The SPA editor runs in your browser. The MCP server integrates with your local AI tools. Your data never leaves your device.
Open source
MIT licensed. Built in the open. Contributions welcome.