get_dashboard_summary
ChatGPTSlim dashboard KPIs and recovery summary. Prefer get_training_status for "what to do today" and get_program_status for per-program progression detail. Host must supply IANA timezone (e.g. Europe/Kyiv) via time_zone or tools/call _meta.clientContext.timezone. Do not ask the human. Do not invent today from raw logs — call this view.
get_exercise_history
ChatGPTRead set-by-set history for one exercise_id (from list_exercises or a program template). For whole-session history use get_workout_history; for program trends use get_training_progress.
get_program
ChatGPTFetch one full program template by program_id (from list_programs). Payload can be large — prefer list_programs when only titles, ids, or revision are needed.
get_program_change_command_schema
ChatGPTReturns the typed contract for propose_program_changes changes[]: every change.type, required/optional fields, workout refs (clientWorkoutKey), scheduleConfig (weeklyDays.dayOfWeek 0–6, daysOfWeek sugar), and worked examples. Use it before composing changes[] if it has not already been retrieved. The schema is stable for the current deployment and can be reused during the conversation; do not refetch unless there is reason to believe the contract changed. It applies to propose_program_changes and is not required for propose_program_creation.
get_program_status
ChatGPTPer-program progression and deload/recovery status (like Calendar program status cards). Use for "how is my progression / recovery week?"; not for picking today's workout (get_training_status) or date-range schedule (get_week_schedule). Host must supply IANA timezone (e.g. Europe/Kyiv) via time_zone or tools/call _meta.clientContext.timezone. Do not ask the human. Do not invent today from raw logs — call this view.
get_training_change_request
ChatGPTRead one user-owned proposal by request_id (status, preview, result). Does not accept, reject, or apply changes — the user does that in the Do Workout app.
get_training_configuration_options
ChatGPTReturn units, defaults, equipment profiles, merged progression profiles, timers, and design limits for building or proposing programs. progressionProfiles includes GZCL Linear Progression (mode gzclp, id f1a2b3c4-d5e6-4789-a012-3456789abc08). Ladder, tier, and increment rules are in get_program_change_command_schema under gzclp. Use include to retrieve only needed sections where practical: equipment, defaults, progressionProfiles, timers, limits. The response versions identify the current profile, progression catalog, and timer catalog inputs. Normally reuse the result during the same interaction; do not repeatedly refetch unless relevant configuration or version information indicates it changed.
get_training_progress
ChatGPTBounded progress aggregates for one program over a date range (sessions, volume, load/e1RM trends, soft signals). Requires program_id, from_date, to_date. Not a substitute for raw history tools or Home "what to do today" (get_training_status). Host must supply IANA timezone (e.g. Europe/Kyiv) via time_zone or tools/call _meta.clientContext.timezone. Do not ask the human. Do not invent today from raw logs — call this view.
get_training_status
ChatGPTAnswer "what should I train today?" — next action and consistency signals (same coaching brain as the app Home). Prefer this over inventing today from get_workout_history. Not the calendar range view (get_week_schedule) or dashboard KPIs (get_dashboard_summary). Host must supply IANA timezone (e.g. Europe/Kyiv) via time_zone or tools/call _meta.clientContext.timezone. Do not ask the human. Do not invent today from raw logs — call this view.
get_user_profile
ChatGPTRead the authenticated user's workout-relevant profile (units, language, progression/equipment defaults, optional body metrics). Use before designing or proposing programs.
get_week_schedule
ChatGPTScheduled workouts for a from_date–to_date calendar range. Use for "what is on my schedule this week?"; not for Home next-action (get_training_status) or progression/deload cards (get_program_status). Host must supply IANA timezone (e.g. Europe/Kyiv) via time_zone or tools/call _meta.clientContext.timezone. Do not ask the human. Do not invent today from raw logs — call this view.
get_workout_history
ChatGPTRead recent completed workout session logs (dates, volume, notes). For "what should I train today?" use get_training_status instead of inventing from history.
list_exercises
ChatGPTSearch the merged Do Workout system + user exercise catalog when exercise IDs are not already available from get_program, prior list_exercises calls, or other Do Workout tool results. Reuse returned IDs; do not re-query merely to verify them because proposal tools validate referenced exercise IDs server-side. Prefer one broad filtered search with an appropriate limit over many single-exercise text searches; paginate with cursor alone when more matches are needed (nextCursor is an offset into the query-matched list). exclude_exercise_ids is optional — it only drops IDs from the returned page after the cursor offset; do not rely on it to walk the catalog. Call this tool sparingly while building a program.
list_programs
ChatGPTList training program summaries for the authenticated user (id, title, workout names, revision, progression flags). Call this first to obtain program_id and revision for get_program or propose_program_changes.
list_training_change_requests
ChatGPTList the authenticated user's AI program proposals (default statuses: pending). Use after propose_* to check status; never claim the program was updated until status is accepted.
propose_program_changes
ChatGPTCreates a pending proposal in Do Workout (TrainingChangeRequest + in-app notification). Does NOT modify the Program. The user must Accept or Reject in the Do Workout app; only Accept applies changes. Never tell the user the program was updated after this tool alone. Requires program_id and expected_revision from list_programs or get_program, plus title, summary, changes[], and idempotency_key (16–128 chars). Referenced catalog/configuration IDs are validated by Do Workout when the proposal is created. After transport uncertainty, retry with the SAME idempotency_key; a new key may create a separate pending proposal. Call get_program_change_command_schema before composing changes[] (all types, fields, clientWorkoutKey workout refs, examples). Each item: { change: { type, ...fields }, reason }. addWorkout: set clientWorkoutKey; reuse it as workoutId / in reorderWorkouts.workoutIds in the SAME proposal. Progression: setProgramProgression = whole-program overwrite (cascades exceptions); setExerciseProgression = per-lift Mixed/exceptions only (assignment none|profile|inherit). GZCLP: setExerciseGzclp sets tier and incrementKg only. Do not send stage, workingWeightKg, cycleStartWeightKg, or consumedLogId.
propose_program_creation
ChatGPTCreates a pending new-program proposal in Do Workout (TrainingChangeRequest + notification). Does NOT create the Program. The user must Accept or Reject in the Do Workout app; only Accept creates it. Never tell the user the program exists after this tool alone. Gather enough context first and reuse exercise IDs from prior Do Workout results. Call list_exercises only for missing IDs, preferring one or a few broad searches; do not make per-exercise verification calls. Requires proposal title, summary, idempotency_key (16–128 chars), and a complete definition shaped as {title, workouts:[{name?, clientWorkoutKey?, exercises:[{exerciseId, prescription?, progression?, gzclp?}], timers?, blocks?}], schedule?, progression?}. For an original GZCLP program use profile f1a2b3c4-d5e6-4789-a012-3456789abc08, four workouts in order A1 squat, B1 press, A2 bench, B2 deadlift, each with T1, T2, and a catalog T3. Base prescriptions are T1 5×3, T2 3×10, T3 3×15. gzclp is {tier, incrementKg}: T1 squat and deadlift 5, T1 bench and press 2.5, every T2 and T3 2.5. Omit schedule. Do not send stage, workingWeightKg, cycleStartWeightKg, or consumedLogId. progression.enabled required when progression is present; exercise progression: assignment inherit|none|profile (+ profileId) — not targetRepRange (profile summary read-only). blocks? follows workoutBlocks in the command schema (omit id). Reject unknown fields (e.g. equipmentProfileId). Validate catalog IDs at propose time. After transport uncertainty, retry with the SAME idempotency_key; a new key may create a separate pending proposal.