Coach Leo — Documentation

Returns the runner's complete coaching context: profile, latest activity, today's planned session, training load trends (ACWR), target race, active body signal alerts, readiness data, recent coaching decisions, and running character status. Must be called at the start of every conversation.

Runner profile and running character — get, update, or recalculate.

• -get_profile — full biometrics, running form data, shoe collection, coaching preferences, training history
• -update_profile — update any profile field including coaching preferences (communication style, training philosophy, intensity preference, rest philosophy, cross-training role, signal alert threshold, fatigue perception bias)
• -get_character — retrieve running character data (requires 7+ days and 5+ activities for eligibility)
• -recalculate_character — recompute character from real activity patterns (called during weekly reviews)

Runner's training activities synced from Strava — search history, get details, record post-workout feedback.

• -query — search activity history with filters: date range, sport type (run, trail_run, ride, mountain_bike, nordic_ski), training impact (primary, cross_aerobic, cross_muscular, cross_recovery)
• -get_detail — full activity details including metrics, heart rate zones, pace splits, and GPX summary
• -record_feedback — record post-workout subjective feedback: perceived effort (1-10), sensations, notes, shoe worn

Training plans and individual sessions — create, modify, add/remove sessions, link completed activities.

• -get_plan — retrieve the active training plan with the next 14 days of scheduled sessions
• -create_plan — build a new structured plan with daily sessions (date, type, title, description, target duration). The knowledge base is always consulted first for training methodology.
• -update_session — mark a session as completed, skipped, or modified, with optional adaptation reason and link to the completed Strava activity

Body signals (tightness, soreness, setbacks) and daily readiness check-ins.

• -record_observation — log training-related feedback: tightness, soreness, asymmetry with body area, severity (info/watch/warning/alert), terrain context, AI interpretation, scientific references, and recommended actions
• -record_physical_setback — record a confirmed physical setback with start date, severity (minor/moderate/severe), type, probable cause, and recovery action
• -record_readiness — daily readiness check-in: sleep hours, sleep quality, energy level, mood, stress level, muscle soreness, resting heart rate

Shoe collection with wear tracking and inspections — add, retire, inspect, link to Strava gear.

• -add — register a new shoe (brand, model, category: trail/road/mixed, estimated lifetime, Strava gear ID)
• -update — modify shoe details
• -retire — mark a shoe as end-of-life
• -inspect — record a zone-by-zone wear inspection (outsole, midsole lateral, midsole medial, upper, heel, insole) with visual score (1-10), observations, AI analysis, and estimated remaining life
• -stats — shoe statistics: total km, inspections history, estimated remaining life
• -link — auto-link shoes to activities via Strava gear IDs

Races from discovery to post-race report — list, register, update, record results.

• -list — get all races by status (planned, completed, DNS, DNF)
• -add — register a race with distance, elevation, terrain type, goal, role (objectif/préparatoire/découverte), and optional parent race for preparation hierarchy
• -update — modify race details
• -report — record race results: finish time, notes, full race report

Scientific knowledge base (training, overtraining, running form, nutrition, trail, psychology). Consulted before every training recommendation, body signal analysis, or running form advice.

• -search — hybrid semantic + full-text search across categories: training methodology, overtraining prevention, trail specific, running form, nutrition & recovery, psychology, cross-training
• -list_sources — browse documents by category and tier (T1 peer-reviewed, T2 expert, personalized)
• -ingest — add a new scientific document with automatic chunking and embedding
• -record_gap — log a missing topic when search returns no relevant results
• -list_gaps — view knowledge gaps by status (open, researching, resolved)
• -resolve_gap — mark a gap as resolved with linked documents

Persistent coaching memory across sessions — log decisions, search past history. Ensures continuity and prevents contradictory advice.

• -log_conversation — save a coaching exchange with topic (onboarding, plan adaptation, body signal analysis, overtraining prevention, weekly review, race prep), summary, decisions, and actions
• -search_history — search past coaching decisions, activities, and training observations by date range and scope

Return current race time projections (5K, 10K, Half-Marathon, Marathon), a trail-adjusted estimate for a target race, and current training zones (Daniels pace zones Z1-Z5 + Karvonen HR zones). Single source of truth shared with the dashboard — same VDOT, same ensemble (VDOT, Tanda, Cameron, Riegel) with dynamic profile-based weighting. Eliminates the risk of the coach improvising race times.

Load a detailed coaching workflow with step-by-step instructions, tool call sequences, and knowledge base queries. Available workflows: onboarding, post-activity analysis, body signal analysis, overtraining monitor, weekly review, adaptive training plan, race planning, deep research, shoe wear analysis, character reveal, arena.