Leo Training

io.github.leo-runningcoachPublished by GitHub account leo-runningcoachv1.1.0Updated Oct 25, 2026

Running, trail and triathlon coach. Reads Strava, Garmin, COROS, Suunto, Polar. Adapts to your feel.

VerifiedStreamable HTTPWeb executableLocation & LifestyleKnowledge & Memory

Overview

AI-generated overview

Hosted remote MCP server for Leo Training, an AI running, trail and triathlon coach that reads your synced training data and builds adaptive plans.

What you can do
Read a full coaching briefing: profile, recent activities, training load, alerts, readiness and running character, plus athlete profile and body signals check-ins. Query and annotate synced activities and record feedback on them. Create, import, modify and track training plans and sessions, sized from real history. Manage races end to end: discovery, registration, preparation, results, debrief, plus race-time and VO2max estimates. Search a 425-document sports-science knowledge base and recall coaching memory across sessions. Track shoes, six-zone wear inspections and retirement, and load multi-step coaching workflows.
When to use it
Use it when you want an assistant to act as a data-driven endurance coach: planning runs, trails or triathlon around your real activities, checking readiness and overtraining signals, preparing a race, or answering training-science questions with cited sources.
Requirements
A remote endpoint at mcp.leotraining.app; nothing to install. A Leo Training account signed in through OAuth 2.1 on the authorization page, and at least one connected training source (Strava, Suunto, Polar, COROS, Garmin, Apple Health or Health Connect). A paid subscription is needed for full coaching.
Before you install
OAuth grants the assistant access to your training and health data, including activities, body metrics and check-ins. Some tools write: plans and sessions are created or modified, activities annotated, and sessions can be pushed to a Suunto watch. A subscription costs money (monthly, quarterly or yearly; a card is required for the 7-day trial). Race discovery uses web search, and data is hosted in Finland.

Installation

In SourceWeft

  1. Open Leo Training in the dashboard and add it to a workspace.
  2. Connect the required credentials. They are stored privately and never shown on public pages.
  3. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.leo-runningcoach%2Fcoachleo/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config. The client completes OAuth sign-in on first connect.

{
  "mcpServers": {
    "leo-training": {
      "type": "http",
      "url": "https://mcp.leotraining.app/mcp"
    }
  }
}

README

[Leo Training]

Leo Training

Leo, your coach for running, trail and triathlon. He reads your real training from Strava, Suunto, Polar, COROS, Garmin, Apple Health or Health Connect, builds your plan from what you actually ran, and adapts it every day to how you actually feel.
On iOS, Android, the web and Telegram, and inside ChatGPT, Claude and Mistral through MCP. Every recommendation is grounded in a knowledge base of 425 sports-science documents graded by level of evidence, across 14 domains.

Website · About · Lab (free science Q&A) · MCP documentation · Smithery

Coach Leo is now Leo Training (since 9 October 2026): same product, same team, same accounts. Leo is still the name of the coach. coachleo.ai and mcp.coachleo.ai keep working during the transition, but new integrations should use leotraining.app and mcp.leotraining.app.

What is Leo Training?

Leo Training is a full coaching application at leotraining.app. It connects to your training data, reads your real history (pace, heart rate, elevation, cadence) and uses specialized coaching tools to make decisions, like a coach who has your training diary open in front of them.

Leo is not a chatbot that writes a generic plan from what you type. He deduces your patterns from your data, recalculates your paces every day from your real runs, and adapts the plan through conversation.

How to get started

  1. Create an account at leotraining.app or download the app — email, Google, or Sign in with Apple
  2. Connect a training source (Strava, Suunto, Polar, COROS, Garmin, Apple Health or Health Connect)
  3. Talk to Leo — he already knows your training, and drafts the skeleton of your plan from your real runs before you enter a card

What makes Leo different

Data first

Leo never asks "what's your weekly mileage?" — he already knows. Strava activities arrive in real time via webhooks; the other sources sync automatically. When you open a conversation, Leo has already read your recent activities, your weekly volume and its trend, your rest days, your plan adherence and your target race.

Paces from your real runs, updated every day

Easy, tempo, interval and long-run paces are derived from your recent training and recalculated every day — no generic templates, no frozen numbers. Your Zone 2 comes from your own heart-rate and pace distributions.

Plans that fit your life

Hard constraints are enforced, not just noted: "I can only run Tuesday, Thursday and Sunday", "40 km a week maximum", "no intervals". When your schedule changes, Leo restructures the plan through conversation while keeping it coherent with your race.

Readiness by feel, not by sensor

Fatigue and readiness come from how you actually feel — a short check-in and a one-tap effort log — not from HRV straps or sleep rings. Most overtraining is invisible to sensors and obvious in a 30-second conversation.

Overtraining prevention on several indicators

Leo never relies on a single metric. He cross-references volume trend, how fast volume rises, rest days, subjective fatigue, recurring body signals and plan adherence. The response is graduated: observation, session change, week restructuring, referral to a professional.

Safety first

Medical and age guards run before any plan is built, and a return-to-run protocol takes over after an injury or an illness.

Running, trail and triathlon

Running comes first, on road and trail. When the goal asks for it, Leo also builds triathlon, duathlon, aquabike and swimrun plans: he sizes the plan on the legs of the race, writes bike sessions (in watts when your FTP is known, by heart rate and effort otherwise) and swim sessions (from your critical swim speed once tested), places brick sessions, and counts one training load across the three sports. He does not project an overall multi-leg finish time.

Race preparation with web search

Say "I'm doing the UTMB CCC" and Leo finds the real race data: date, distance, elevation, terrain, GPX course. He analyzes the profile against your training and estimates your finish time with several models (a physiological endurance index, Riegel, Cameron, Tanda), trail-adjusted for elevation.

Race-time estimates without a race

Leo keeps estimating your race times across distances from training data alone, and tracks them over time so you can see your fitness trend.

Sports nutrition

Evidence-based race fueling adapted to your weight, duration, elevation and gut tolerance, following IOC/ISSN guidelines, plus recovery nutrition timing and carbohydrate periodization.

A scientific knowledge base, graded by level of evidence

425 documents (consensus statements, meta-analyses, systematic reviews, controlled trials, cohort studies) across 14 domains: training methodology, running form and biomechanics, injury prevention, medical safety, nutrition, mental performance, physiology, runner populations, cross-training, sleep and recovery, environmental factors, load management, trail running and multisport. Each document carries its level of evidence and status, and search ranks a recent, better-supported recommendation above an older or weaker one. New documents are added weekly, after a human check of the consensus, and only from legal open-access texts.

Try it free, with no account, at leotraining.app/lab: ask any running science question and get strictly cited answers.

Persistent coaching memory

Leo searches past history before advising and references specific conversations: "Last week you mentioned calf tightness and your volume is up 30% — let's be careful." After three months, Leo holds a coaching record no new app can replicate.

Shoe wear analysis

Not just mileage. Leo inspects six wear zones (outsole, lateral midsole, medial midsole, upper, heel, insole). Wear patterns reveal biomechanical signals, which feed back into body-signal tracking.

VO2max and training zones

VO2max cross-validated from several methods with confidence intervals; pace and heart-rate zones derived from your physiology, with cycling power zones when an FTP is set.

Running characters

Nine animal archetypes (Wolf, Mountain Goat, Turtle, Phoenix, Hawk, Bear, Fox, Cheetah, Otter) computed from nine real training dimensions — not a personality quiz. They evolve through three stages and shape how Leo talks to you.

Cross-training with a purpose

Cycling, swimming, hiking, skiing: each session gets a role (recovery, aerobic development, muscular reinforcement, injury prevention) and counts in your weekly load.

Connect your data

All sources connect from one hub in the app.

SourceStatusWhat it does
StravaLiveRead-only sync, activities arrive in real time via webhooks.
SuuntoLiveReads your watch activities, and pushes structured sessions back to the watch via SuuntoPlus Guides.
PolarLiveReads your activities via Polar AccessLink (read-only).
COROSLiveReads your activities and their laps, with six months of history on first connection.
GarminLiveReads your Garmin Connect activities.
Apple Health / Health ConnectLiveImports workout history through the native iOS and Android apps.

Available everywhere

ChannelDescription
iOS & AndroidNative apps on the App Store and Google Play, with push notifications and health-data import.
Web appChat, training dashboard, character, race countdown and performance metrics at leotraining.app. Installable as a PWA.
TelegramMessage Leo from your everyday messaging app. Same memory, same tools, same coach.
ChatGPT, Claude, MistralConnect Leo through MCP and your assistant gets full access to your training. Leo is listed in the official ChatGPT plugin directory.

Community: the Arena

  • Season 02, "The Ekiden Road" — seven stages across a Japan of legend, run as one collective relay: every session a runner logs moves the road forward.
  • Solo quests — personal challenges tied to your training dimensions and your character.
  • Social feed — built on the grammar of process, not chrono-chasing.

The Arena is not a competition: no ranking, nobody sees what you missed, and the community crosses each stage together.

Pricing

One subscription with all the coaching included. The amount is the same in USD, EUR, GBP and CHF.

  • Monthly: 9.99 per month (founding-member rate; standard rate 12.99)
  • Quarterly: 26.99 per quarter (8.99 per month)
  • Yearly: 79.99 per year (6.67 per month)

Every new subscription starts with 7 free days: a card is required, nothing is charged before day seven, and cancelling takes one tap. The 7 free days are granted once per account; a subscription that starts without them is covered by a 14-day money-back guarantee. Creating an account, connecting a source and the first conversation with Leo cost nothing.

On iOS, subscriptions go through Apple In-App Purchase; web and Android use Stripe with a self-service portal.

MCP server (for developers)

Leo Training is a hosted remote MCP server — nothing to install. Connect it to any MCP-compatible client (ChatGPT, Claude, Mistral, Cursor, Windsurf and more).

Connecting is open to any signed-in runner. A new account is offered the 7-day trial on the authorization page, once, when it links the assistant; after that the conversation never stops to ask for payment, and no purchase link is ever sent inside the assistant.

Quick setup

  1. Open your MCP client's settings
  2. Add a remote server: https://mcp.leotraining.app/mcp
  3. Sign in with your Leo Training account and approve the connection
  4. Start talking — the coaching tools are ready

Step-by-step guides: ChatGPT · Claude (custom connector, works on every plan) · other clients

JSON config

json
{  "mcpServers": {    "leo-training": {      "url": "https://mcp.leotraining.app/mcp"    }  }}

Coaching tools

ToolDescription
get_coaching_briefingFull coaching context: profile, recent activities, training load, alerts, readiness, character.
athleteRunner profile and character: body metrics, preferences, character recalculation.
activityQuery and annotate synced activities, record feedback.
trainingTraining plans and sessions: create, modify, track adherence.
create_my_planBuilds a training plan from a short brief (goal, race, date); the server sizes it from the runner's real history.
register_planImports the plan a runner already follows (own plan, club coach) so Leo can track and adapt it.
body_signalsBody observations, setbacks and daily readiness check-ins.
shoesShoe collection, six-zone wear inspections, mileage and retirement.
raceFull race lifecycle: discovery via web search, registration, preparation, results, debrief.
race_performanceRace-time estimates and VO2max from training data, trail-adjusted, tracked over time.
knowledgeHybrid semantic and full-text search across the 425-document knowledge base.
memoryCoaching memory across sessions: decisions, context, consistency.
get_skill_workflowLoads a coaching workflow that chains tools for a complex scenario.

Clients that support MCP Apps also get visual cards: today's home view, plan week and block, session analysis, target race, race projection and training journal.

Coaching workflows

  • Onboarding — data first: Leo deduces your patterns before asking anything
  • Post-activity analysis — debrief with shoe deduction and sensation logging
  • Adherence & feeling check — quick readiness and adherence pulse
  • Body signal analysis — injury-risk assessment backed by cited sources
  • Overtraining monitor — multi-indicator load analysis with graduated alerts
  • Weekly review — training summary, character update, plan adjustments
  • Adaptive training plan — plans from real paces with constraint validation
  • Race planning — race data from the web, GPX analysis, taper
  • Post-race debrief and post-race recovery — result analysis and a safe return to training
  • Deep research — knowledge-base search on a specific question
  • Shoe wear analysis — six-zone inspection with biomechanical patterns
  • Character reveal — running character assignment
  • Arena — quests and the collective season

Authentication

Full OAuth 2.1:

  • Discovery: https://mcp.leotraining.app/.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
  • Training sources are read-only (except sessions you choose to push to your watch)
  • Strict per-user data isolation
  • GDPR compliant — export and deletion on request
  • Your data is never used to train AI models

Privacy policy · Terms · Refund policy

Links

Made in the Alps.

Source: README.md at commit 4e9e570

Tools

0
This server asks you to sign in, so SourceWeft could not read its tools without an account and has not indexed them. What a connected account sees can depend on that account and its permissions.

Version history

2
  1. v1.1.0LatestOct 14, 2026
  2. v1.0.1Sep 16, 2026

Score

75% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.