Coach Leo — AI Running Coach
An AI running coach that reads your real training from Strava, Suunto, Polar, Apple Health or Health Connect, builds plans from your actual history, and adapts daily to how you actually feel.
Available on iOS, Android, on the web, on Telegram, and as an MCP server for ChatGPT, Claude, and Mistral. Every recommendation draws from 250+ peer-reviewed sports science papers across 13 research domains.
Try it free ·
About ·
Lab (free science Q&A) ·
MCP Documentation ·
Smithery
What is Coach Leo?
Coach Leo is a full AI running coach application at coachleo.ai. It connects directly to Strava, Suunto or Polar — or imports from Apple Health on iOS and Health Connect on Android — reads your real training history (pace, heart rate, elevation, cadence), and uses specialized coaching tools to make decisions, like a real coach with access to your training diary.
Leo is not a chatbot that generates generic plans from what you type. He deduces your training patterns from your data, derives your actual paces from 8 weeks of history, and adapts your plan daily through conversation.
How to get started
- Create an account at coachleo.ai or download the application — email, Google, or Sign in with Apple
- Connect your Strava account (read-only)
- Start chatting — Leo already knows your training
What makes Leo different
Data-first coaching
Leo never asks "what's your weekly mileage?" — he already knows. Strava activities arrive in real time via webhooks, within seconds of your watch syncing. When you open a conversation, Leo has already analyzed your last 10 activities, your training load, your acute-to-chronic workload ratio (ACWR), and your weekly patterns.
Plans built from YOUR actual paces
Leo derives your easy, tempo, interval, and long-run paces from 8 weeks of real training data. No generic templates. Your Zone 2 pace comes from YOUR heart rate and pace distributions.
Plans that fit your life
Hard constraints are enforced, not just preferences. "I can only run Tuesday, Thursday, and Sunday", "maximum 40km per week", "no intervals" — Leo respects these in every plan. When your schedule changes, Leo restructures through conversation while keeping it coherent with your race goal.
Readiness by feel, not by sensor
Fatigue and readiness are read from how you actually feel, through conversation and a one-tap effort log — not from HRV straps or sleep wearables. A short check-in is enough for Leo to adjust today's session.
Multi-indicator overtraining prevention
Leo never relies on a single metric. He cross-references ACWR trends, volume increase speed, rest days, subjective fatigue, recurring body signal patterns, and plan adherence. Response is graduated: observation, session modification, week restructuring, professional referral.
Biomechanical shoe wear analysis
Not just mileage tracking. Leo inspects 6 wear zones (outsole, lateral midsole, medial midsole, upper, heel, insole) with visual scoring. Wear patterns reveal biomechanical signals — pronounced medial wear suggests overpronation, which feeds back into body signal tracking.
Race preparation with web search
Say "I'm doing the UTMB CCC" and Leo searches the web for real race data: date, distance, elevation, terrain, participants, GPX file. No manual entry. Leo then analyzes the GPX profile and predicts your finish time using multiple models (VDOT, Riegel, Cameron, Tanda).
Race-time estimates without a race
Leo continuously estimates your race times across distances from training data alone — no recent race required. Estimates are VDOT-anchored, trail-adjusted for elevation, and tracked over time so you can see your fitness trend.
Sports nutrition guidance
Evidence-based race fueling adapted to your weight, duration, elevation, and gut tolerance. IOC/ISSN guidelines. Recovery nutrition timing and carb periodization. Per-session nutrition plans included.
Open scientific knowledge base
250+ peer-reviewed papers across 13 categories (training methods, physiology, running form and biomechanics, injury prevention, nutrition, recovery, sleep, strength training, mental performance, race strategy, environmental factors, and trail-specific training) with hybrid semantic search (BM25 + pgvector). Try it free, no account, at coachleo.ai/lab — ask any running science question and get strictly cited answers.
Persistent coaching memory
Leo searches past history before advising. He references specific conversations: "Last week you reported calf tightness and your ACWR was 1.4 — volume is up 30%, let's be cautious." After 3 months, Leo has a coaching dossier no new app can replicate.
VO2max and fitness predictions
VO2max cross-validated from multiple scientific methods (ACSM, Jack Daniels VDOT, HR-Pace regression) with confidence intervals. Pace and HR training zones derived from your physiology, with cycling power zones when an FTP is set.
Running character system
9 animal archetypes (Wolf, Mountain Goat, Turtle, Phoenix, Hawk, Bear, Fox, Cheetah, Otter) computed from real training dimensions — not a personality quiz. Calculated from your data. Evolves through 3 stages. Your character shapes how Leo coaches you.
Cross-training as strategic lever
Cycling, swimming, hiking, skiing — each session gets a specific role: recovery, aerobic development, muscular reinforcement, or injury prevention. Integrated into weekly training load.
Connect your data
Coach Leo is built to reduce single-source dependence. All sources connect from one unified hub at coachleo.ai/connect.
| Source | Status | What it does |
|---|
| Strava | Live | Primary, read-only sync — activities arrive in real time via webhooks. |
| Suunto | Live | Read your watch activities — and push structured sessions back to the watch via SuuntoPlus Guides. |
| Polar | Live | Read your activities via Polar AccessLink (read-only). |
| Apple Health / Health Connect | Live | Import workout history through the native iOS and Android apps to complement your main source. |
| Garmin · Coros · .fit | Coming soon | More watch ecosystems and file import. |
Available everywhere
| Channel | Description |
|---|
| Web app | Full chat interface with training dashboard, character visualization, race countdown, and performance metrics at coachleo.ai |
| Telegram | Message Leo directly for quick coaching on the go. Same memory, same tools, same coach. |
| ChatGPT, Claude, Mistral | Connect Leo as an MCP server. Your favorite AI assistant becomes your running coach with full access to your data. |
| iOS & Android | Native apps on the App Store and Google Play — device health import and push notifications. |
Train alongside runners who share your training DNA.
- Packs — Your character determines your pack. Join runners worldwide who share your training DNA, or create an elective pack with an invite code (public or private).
- Solo quests — Personal challenges tied to your training dimensions and your character.
- Pack battles — Weekly and seasonal competitions between packs.
- Expeditions — Run together across iconic routes worldwide. Every kilometer counts toward the collective journey.
- Social feed — A feed built on the grammar of process, not chrono-chasing — celebrate the work, not just the times.
Activity tracking
Coach Leo syncs all your activities — running is the primary coaching focus (road, trail, track), but every other sport you log (cycling, swimming, CrossFit, hiking, skiing, etc.) is tracked with a specific role to give the coach a complete picture of your training load and recovery needs.
Pricing
- Free plan — start with no card: a personalized starter plan, your running character, and community access.
- Founding member: $9.99/month (limited spots, locked for life — regular price is $12.99/month)
- Quarterly: $26.99 every 3 months
- Annual: $79.99/year (≈ $6.67/month)
- Paid unlocks the full adaptive coach: plans that adapt to your data, unlimited chat, premium analysis, and the MCP server.
On iOS, subscriptions are handled through Apple In-App Purchase; web and Android use Stripe with a self-service portal.
MCP Server (for developers)
Coach Leo also works as a hosted remote MCP server — no installation required. Connect it to any MCP-compatible client (ChatGPT, Claude Desktop, Cursor, Windsurf, and more).
Connecting the Coach Leo MCP is reserved for subscribers (like a Strava-grade integration). Existing tokens keep working; a new connection prompts you to subscribe.
Quick setup
- Open your MCP client's settings
- Add a new remote server:
https://mcp.coachleo.ai/mcp
- Authenticate with your Coach Leo account
- Start chatting — the coaching tools are ready
JSON config
{
"mcpServers": {
"coach-leo": {
"url": "https://mcp.coachleo.ai/mcp"
}
}
}
| Tool | Description |
|---|
get_coaching_briefing | Complete coaching context: profile, recent activities, training load, alerts, readiness, character status. |
athlete | Runner profile and running character management. Body metrics, preferences, character recalculation. |
activity | Query and annotate synced activities. Filter by date, sport, impact. Record feedback. |
training | Create and manage personalized training plans. Modify sessions, track adherence. |
body_signals | Log body observations, record physical setbacks, daily readiness check-ins. |
shoes | Track shoe collection, wear inspections across 6 zones, mileage stats, and retirement. |
race | Full race lifecycle: discovery via web search, registration, preparation, results, post-race analysis. |
race_performance | Race-time estimates and VO2max across distances from training data — trail-adjusted, tracked over time. |
knowledge | Hybrid semantic + full-text search across 250+ peer-reviewed papers. |
memory | Persistent memory across sessions. Logs decisions, prevents contradictory advice. |
get_skill_workflow | Loads a predefined coaching workflow that chains tools for a complex scenario. |
Coaching workflows
Predefined workflows that chain tools together for complex scenarios:
- Onboarding — Data-first setup: Leo deduces your patterns before asking anything
- Post-activity analysis — Automated debrief with shoe deduction and sensation logging
- Adherence & feeling check — Quick readiness and plan-adherence pulse
- Body signal analysis — Injury risk assessment backed by knowledge base citations
- Overtraining monitoring — Multi-indicator load analysis with graduated alerts
- Weekly review — Training summary with character recalculation and plan adjustments
- Adaptive plan creation — Plans from real paces with constraint validation
- Race planning — Web search for race data, GPX analysis, taper protocol
- Post-race debrief — Result analysis and lessons for the next block
- Deep research — Knowledge base enrichment from web search
- Shoe wear assessment — 6-zone inspection with biomechanical pattern detection
- Character reveal — Running character assignment ceremony
- Arena — Packs, quests, and expeditions
Authentication
Full OAuth 2.1 implementation:
- Discovery:
/.well-known/oauth-authorization-server
- Dynamic Client Registration (DCR)
- PKCE with S256
- Token refresh and revocation
Privacy & data
- Hosted in Finland (Hetzner, EU)
- HTTPS/TLS encryption
- Strava data is read-only
- Strict user data isolation
- GDPR compliant — export and deletion on request
- No data used for AI model training
Links
Made in the Alps.