cancel_plan_job
ChatGPTRequest cancellation of an owned queued or running plan job. The current provider request is allowed to finish; processing stops before the next item. Completed creates or deletes are never rolled back.
create_training_plan
ChatGPTValidate and format a training plan of 1–180 day-numbered workout rows in TrainingDojo's standard format. Returns per-row validation errors, or a preview with calculated calendar dates and plan totals for the user to review. Does NOT push to any platform; push_training_plan publishes a plan with the same fields.
delete_plan
ChatGPTEnqueue durable deletion of every remaining platform workout in a tracked plan and return immediately. Provider not-found responses count as successful idempotent deletion. Completed deletes are removed from the tracked plan as progress is made; the plan record is removed only when no tracked workouts remain. This is destructive. The connected AI provider controls whether it shows an approval prompt.
delete_platform_workout
ChatGPTDelete a planned workout from TrainingPeaks or Intervals.icu, dropping the session entirely (update_platform_workout modifies a session instead). Workout IDs come from get_upcoming_workouts. This permanently removes the workout from the platform calendar. TrainingPeaks coaches must pass the same athlete_id used to read the workout.
get_activity_detail
ChatGPTFetch the full detail of a single completed activity or workout — the "how did I do in that session?" tool. Get activity IDs from list_recent_activities (or workout IDs from get_upcoming_workouts for TrainingPeaks). Intervals.icu can also include the per-interval breakdown (power/HR per interval) via include_intervals. TrainingPeaks coaches pass the same athlete_id used to find the workout.
get_athlete_profile
ChatGPTReturns the athlete's profile including body weight (kg) and workout builder preferences (warmup/cooldown duration and intensity ranges, recovery settings). Use this alongside training history to set appropriate training zones when building a plan. Workout builder preferences help the connected AI client build matching structured workout JSON.
get_connected_platforms
ChatGPTReturns the training platforms the user has connected to TrainingDojo (TrainingPeaks, Intervals.icu), including per-platform capabilities (pushStructured, pushPlan, calendarRead, activityList, activityDetail, wellness, updateWorkout, deleteWorkout). These are the data sources and push destinations available to every other tool. If no platforms are connected, the result says so; platforms are connected at https://trainingdojo.app/mcp.
get_intervals_history
ChatGPTFetch and analyze the athlete's recent Intervals.icu activity history. Returns volume totals, weekly averages, sport mix, performance trend (CTL/ATL/TSB from wellness data or estimated from load), consistency metrics, and long-session patterns. Requires Intervals.icu to be connected at https://trainingdojo.app/mcp.
get_plan_job
ChatGPTReturn an owned publish/delete job and a paginated safe item summary. When job_id is omitted, returns the most recently updated owned job. Encrypted workout payloads are never returned.
get_trainingpeaks_compliance_report
ChatGPTBuild a deterministic weekly compliance report for all or selected athletes on the connected TrainingPeaks coach account (an athlete account reports on itself). For each planned session it returns TrainingPeaks' reported values, a completion status, and separate actual/planned ratios for duration, TSS, and distance; there is no blended score. Each session has a derived.label, and a definitions block explains every field. A session with no completion recorded is reported as no_completion_recorded, separate from skipped, because its activity may not be uploaded yet. Processes up to 10 athletes per call and returns next_cursor until report.complete is true. Athlete IDs and the coach time zone come from list_trainingpeaks_athletes; week_start is the first day of the report week and as_of is the current date in that time zone.
get_trainingpeaks_history
ChatGPTFetch and analyze the athlete's recent TrainingPeaks workout history. Returns volume totals, weekly averages, sport mix, performance trend (CTL/ATL/TSB from TP performance data), consistency metrics, and long-session patterns. Use this to understand the athlete's actual training load before building a plan. TrainingPeaks coaches can pass athlete_id to analyze one attached athlete. Requires TrainingPeaks to be connected at https://trainingdojo.app/mcp.
get_upcoming_workouts
ChatGPTRead planned and completed workouts from the athlete's TrainingPeaks or Intervals.icu calendar for a date range, past or future. TrainingPeaks rows include planned values, recorded actual values, and TrainingPeaks' own compliance percentages when present; missing values are null, never zero. Use this to see what is scheduled before adapting a plan, to review a past week, and to fetch platform workout/event IDs for update_platform_workout / delete_platform_workout. Defaults to the next 14 days anchored to UTC. For precise athlete-local calendar boundaries, pass explicit from_date and to_date values. TrainingPeaks coaches can pass athlete_id to read one attached athlete; hiddenFromAthlete marks drafts the athlete cannot see yet.
get_vault_workout
ChatGPTFetch a single Workout Vault row including its generated structured workout (canonical JSON) when available. Use the canonical structure directly with push_structured_workout, or publish it via publish_vault_workout.
get_wellness
ChatGPTFetch the athlete's wellness metrics for a date range. Intervals.icu returns daily wellness records (CTL/ATL fitness/fatigue, plus sleep, HRV, resting HR, weight and anything else the athlete logs). TrainingPeaks returns logged metrics (Weight Kilograms, Fatigue, HRV, Sleep Hours, ...). Use this before adapting a plan — e.g. poor sleep or depressed HRV argues for easing the next sessions. Defaults to the last 14 days anchored to UTC. For precise athlete-local calendar boundaries, pass explicit from_date and to_date values. TrainingPeaks coaches can pass athlete_id for one attached athlete.
get_workout_types
ChatGPTReturns the available sport types and subtypes for a platform. Use this to understand valid sport/subtype values before calling create_training_plan or push_training_plan.
list_plans
ChatGPTList training plans previously imported to TrainingPeaks or Intervals.icu through TrainingDojo (including plans pushed via push_training_plan). Returns plan IDs for delete_plan. Note: only plans imported through TrainingDojo are tracked here — use get_upcoming_workouts to see everything on the platform calendar.
list_recent_activities
ChatGPTList the athlete's recent completed activities with summary metrics (date, name, sport, duration, distance, TSS/load, avg power/HR). Use this to find a specific session to discuss, then call get_activity_detail with its activity ID for the full breakdown. Defaults to the last 14 days, with calendar boundaries anchored to UTC; near the athlete's local midnight, account for that offset. TrainingPeaks coaches can pass athlete_id to list one attached athlete's activities.
list_trainingpeaks_athletes
ChatGPTList the athletes attached to the connected TrainingPeaks coach account, with the IDs to pass as athlete_id to other tools (calendar, activities, detail, wellness, history, updates, deletes, and the weekly compliance report). The coach's own account ID is not a valid athlete target. For an athlete account, returns only that athlete. Also returns the coach's TrainingPeaks time zone (use it to choose week_start and as_of dates) and TrainingPeaks' compliance metric priority. Names only; no contact details.
list_vault_workouts
ChatGPTList the user's Workout Vault — their library of saved workouts imported from training history and converted to reusable structured workouts. Returns row IDs for get_vault_workout and publish_vault_workout. Defaults to the most recent vault batch; pass batch_id to browse older batches (also listed in the response).
publish_vault_workout
ChatGPTPush a generated Workout Vault workout to the athlete's TrainingPeaks or Intervals.icu calendar on a given date. Does not invoke TrainingDojo-hosted AI because the structure was already generated. The row must have a generated structure (see list_vault_workouts hasStructure).
push_structured_workout
ChatGPTPush a structured workout to TrainingPeaks or Intervals.icu without invoking TrainingDojo-hosted AI. Build the workout JSON yourself (the input schema documents the exact structure: flat intensityMin/intensityMax as percentages for FTP/threshold metrics or raw 1-10 values for rpe, lengths in seconds or meters, repetition blocks for intervals). The connected AI client builds this JSON using its own model subscription. Tip: call validate_structured_workout first to catch structural problems and preview duration/TSS. Requires the platform to be connected at https://trainingdojo.app/mcp.
push_structured_workouts_bulk
ChatGPTPush multiple structured workouts to TrainingPeaks or Intervals.icu without invoking TrainingDojo-hosted AI. Processes sequentially at 1.5 req/sec and reports per-row succeeded, failed, unknown, or unattempted outcomes. Row-specific failures and unknown outcomes do not discard earlier successes; connection-wide failures stop later rows. Each workout uses the same structure as push_structured_workout; call validate_structured_workout on drafts first to catch structural problems. Retry only failed or unattempted rows with a new key; verify unknown rows on the provider calendar first.
push_training_plan
ChatGPTValidate and enqueue 1–180 simple calendar workouts for durable best-effort publishing to TrainingPeaks or Intervals.icu. Returns immediately with a job ID. Use get_plan_job to inspect progress, cancel_plan_job to stop before the next provider request, and retry_plan_job for eligible rows. The caller-provided UUID makes repeat calls replay the exact result instead of republishing successful rows.
retry_plan_job
ChatGPTRetry only failed, canceled, or unattempted rows in an owned terminal job. Successful publish rows are never republished. Unknown publish outcomes require include_unknown=true after the user reviews the risk; unknown deletes may be retried safely. Returns the existing job when no row is eligible.
update_athlete_profile
ChatGPTPersist athlete profile settings the coaching conversation surfaces: body weight (kg) and workout-builder preferences (warmup/cooldown duration and intensity, recovery settings). These preferences guide structured workout JSON and are shown in get_athlete_profile. Only the fields you pass are changed.
update_platform_workout
ChatGPTUpdate a planned workout on TrainingPeaks or Intervals.icu — move it to another day, retitle it, or change its planned duration/TSS/description. This is the adaptive-coaching tool: use it to adjust upcoming sessions after discussing recovery or performance. Fetch IDs with get_upcoming_workouts first. Only the fields you pass are changed; everything else is preserved. TrainingPeaks coaches must pass the same athlete_id used to read the workout, and can pass hidden_from_athlete to hide a workout or reveal a hidden draft.
validate_structured_workout
ChatGPTValidate a structured workout JSON draft before pushing. This is deterministic, invokes no hosted AI, and has no side effects. Returns per-field issues (with JSON paths) when invalid, or a summary (step count, total duration, estimated IF and TSS for complete duration-based percentage workouts) when valid. Iterate with this tool until valid, then call push_structured_workout. Workout shape: {primaryLengthMetric:"duration"|"distance", primaryIntensityMetric:"percentOfFtp"|"percentOfThresholdHr"|"percentOfThresholdPace"|"rpe", steps:[...]} where each step is {type:"step", name, length:{value, unit:"second"|"meter"}, intensityMin, intensityMax (percentage, or raw 1-10 for rpe), intensityClass:"warmUp"|"active"|"rest"|"coolDown"|"other", cadenceMin?, cadenceMax?} and interval sets are {type:"repetition", name, reps, steps:[step,...]}. Ramps: {type:"rampUp"|"rampDown", name, steps:[2+ steps]}. Distance steps in duration-axis workouts need an honest estimatedDurationSeconds; timed steps are valid on a distance axis. Any distance step requires top-level visualizationDistanceUnit ("meter"|"kilometer"|"mile"|"yard").
cancel_plan_job
Claudecreate_training_plan
Claudedelete_plan
Claudedelete_platform_workout
Claudeget_activity_detail
Claudeget_athlete_profile
Claudeget_connected_platforms
Claudeget_intervals_history
Claudeget_plan_job
Claudeget_trainingpeaks_compliance_report
Claudeget_trainingpeaks_history
Claudeget_upcoming_workouts
Claudeget_vault_workout
Claudeget_wellness
Claudeget_workout_types
Claudelist_plans
Claudelist_recent_activities
Claudelist_trainingpeaks_athletes
Claudelist_vault_workouts
Claudepublish_vault_workout
Claudepush_structured_workout
Claudepush_structured_workouts_bulk
Claudepush_training_plan
Clauderetry_plan_job
Claudeupdate_athlete_profile
Claudeupdate_platform_workout
Claudevalidate_structured_workout
Claude