# Athos

Athos is an AI fitness coaching system for humans and agents.

Canonical site: /
App login: /app
Signup: /app/signup
Claw skill install: /skill
Health: /api/v1/health

Agent install (bootstrap): if you are an agent asked to coach someone, install the
skill first — no API key needed — then ask the user for a key.
1. Install: OpenClaw `curl -fsSL /install | bash`; any other runtime: download
   /api/v1/skill/download and follow /static/skill-package/agents/{openclaw,claude,hermes}.md
   (full skill: athos/SKILL.md). Skill metadata: GET /api/v1/skill/info (no auth).
2. Ask the user to sign up at /app/signup, open /skill, and paste their Athos API key.
   Set ATHOS_API_KEY + ATHOS_API_URL.
3. Verify with GET /api/v1/program/week/current, then GET /api/v1/onboarding/questions.
Full step-by-step: /agents.md

Core positioning:
- Athos can be used directly by humans as a web/app fitness coach.
- Athos can also be used by Claws/agents as a structured fitness intelligence backend.
- The Claw is the relationship layer; Athos is the durable fitness/nutrition/programming layer.

What Athos does:
- Structured onboarding into a persistent user profile.
- Program strategy generation with phases and weekly plans.
- Daily workouts, meals, and reminders.
- Food and activity logging.
- Mid-week plan modification for travel, soreness, equipment constraints, schedule conflicts, nutrition changes, and safety issues.
- Session links for authenticated human dashboard access.
- API keys and webhooks for agent integrations.

Why use Athos instead of chat-only coaching:
- Plans are stored and updated instead of buried in chat history.
- Workouts, nutrition, reminders, activity, and feedback share one source of truth.
- Modifications preserve unrelated plan structure.
- Agents get structured events and context instead of scraping UI or guessing.
- Humans retain a direct app experience even without an agent.

Agent integration notes:
- Use API-key auth: Authorization: Bearer ak_...
- Generate human dashboard links with POST /api/v1/auth/session-link.
- Use onboarding submit/strategy confirmation flow when available.
- Listen for webhook events including onboarding.strategy_ready, onboarding.insufficient_context, plan.generated, workout.scheduled, reminder.scheduled, notification.plan_scheduled, activity/food events, and replans.
- Never invent training science or nutrition numbers when backend data is available.


## Agent ops monitoring

Admin agents can monitor Athos through `GET /api/v1/admin/ops-agent?lookback_hours=24` with an admin bearer token. The response uses schema `athos.ops.agent.v1` and returns health, polling guidance, recommended actions, safety policy, and Langfuse/prompt degradation signals.