MCP App Store
by KeenEthics
wolf icon

wolf MCP Server for ChatGPT

By CHRISTOPHER DANA FLEISHER

Overview

Review your fitness goals, workouts, and recovery trends, and plan training using data from your connected wolf account. Designed for fitness and general wellness; does not provide medical diagnosis or treatment. A wolf account is required.

How to connect

  1. ChatGPT: Open the wolf listing in the ChatGPT app directory and choose Connect. ChatGPT listing wolf exposes 39 tools here.

Tools

abandon_plan

ChatGPT
Abandon a whole week (proposed or active) by plan_id: a duplicate standalone week, or a proposal the athlete should not approve. Nothing is deleted; the week stops showing. To drop a single session use remove_plan_session.

add_sessions_to_plan

ChatGPT
Add sessions to an existing week (a cycle week or a standalone week) as a new version. sessions: [{"day": mon..sun, ...session as in save_weekly_plan}]. An active week stays active, so the additions are live immediately; a proposed week stays proposed. plan_id comes from get_schedule, get_cycles or get_plan_compliance.

adopt_sessions_into_cycle

ChatGPT
Fold previously independent (standalone) sessions into a cycle: every standalone session on each date in dates (YYYY-MM-DD) moves into the cycle's week for that date. The cycle week must already exist (save_weekly_plan); it keeps its status. The standalone week loses those sessions (abandoned if none remain).

annotate_workout

ChatGPT
Attach a short note (author 'mcp') to a workout, e.g. an observation about drift or pacing.

create_exercise

ChatGPT
Add a new exercise to the catalogue when resolve_exercises finds no match: name, movement (push, pull, squat, hinge, carry, core, cardio — you decide, the athlete never picks one), optional aliases and unilateral flag. Refused (409) when the name or an alias is already in the catalogue as written or up to equipment words, so history never splits; a merely similar name is accepted. VARIANTS: a tweak of an existing exercise (pause, tempo, close grip, deficit …) is a variant, not a new exercise: pass parent (the primary's name) and variant (the short label); movement is inherited and the name should be like 'Bench press (pause)'. The primary's history then shows the variant as its own series. Names like 'paused bench press' in log_performed / plans create the variant automatically when create_missing is on, but only when the rest of the name is the primary's own name; otherwise the name becomes a standalone exercise. resolve_exercises reports inferred_parent/inferred_variant so you can decide first.

detach_session_from_cycle

ChatGPT
Take one session (by day and title) out of a cycle week and keep it on the same date as a standalone session. Returns the cycle week and the standalone plan.

get_coaching_context

ChatGPT
Read the athlete's EXISTING goals/targets, cycles, outlined sessions and recent workouts before coaching, setup or making a proposal. Pure read: never initializes or replaces anything. Returns saved IDs/statuses, missing areas, and coaching guidance. A read failure must not be interpreted as an empty account. tz: IANA zone; optional workout_id selects a recorded session belonging to this athlete. Health data must be genuine synced HealthKit readings; fabricated sample values and medical diagnosis are unsupported, including for test accounts.

get_coaching_plan_state

ChatGPT
Refresh saved cycle weeks, weekly proposals and schedule after a decision.

get_cycles

ChatGPT
Training cycles with their weeks, per-week compliance, the load ramp (template vs planned vs actual TRIMP) and guardrail flags (ACWR ceiling, minimum easy share). Proposed and active by default; include_inactive adds finished and superseded ones.

get_exercise_history

ChatGPT
One exercise's history by catalogue name or alias (e.g. 'bench', 'kb swing'): rep maxes 1-12 (best kg at >= n reps, with date), estimated 1RM per session (Epley), session volume, every loaded set as a load-by-reps point with its rep band, and recent sessions. Loads are kilograms.

get_fitted_models

ChatGPT
The athlete's fitted personal models (ADR 0003): today the hill curve, the grade-adjustment multipliers fitted from their own runs at matched heart rate, per grade bin with stretch counts, 95 % bands, trust, the held-out backtest against the population (Minetti) curve, and whether the app is using the personal or the population curve for grade-adjusted pace. Cite it when you talk about hills, GAP or race pacing.

get_goals

ChatGPT
The athlete's goal sets (with weighted roll-up: progress, completion, trajectory, quality, excluded_weight) and standalone goals, each with today's evaluation (current, progress, expected, trajectory, quality). Active by default; include_inactive adds proposals awaiting approval, achieved / missed history and superseded versions. tz is the athlete's IANA zone. Units are SI: vo2max_latest mL/kg/min, rhr_7d bpm, hrv_7d ms, sleep_h hours, weekly_km km per week, weekly_min minutes per week, body_mass kg, exercise_rm kg (spec.exercise + spec.reps). Dates are YYYY-MM-DD.

get_ingest_status

ChatGPT
Per-HealthKit-type sample counts and date ranges, plus workout totals: how much data exists and how fresh it is.

get_plan_compliance

ChatGPT
Compliance of one planned week: each session done / partial / missed with the matched workout, work-step time-in-range for structured runs, logged-set checks for strength, extra workouts, and planned vs actual minutes, load and easy share. Identify the week by week_start (Monday, YYYY-MM-DD) or plan_id; without either, the current week. tz is the athlete's IANA zone.

get_schedule

ChatGPT
The athlete's forward (or past) schedule: every non-rest session of every active plan between from and to (YYYY-MM-DD; default today through 13 days ahead), cycle weeks and standalone suggestions merged, each with date, plan_id, cycle_id, the session and its compliance result once the day has passed. Use plan_id, day and title from here with add_sessions_to_plan / replace_plan_session / remove_plan_session.

get_strength_volume

ChatGPT
Strength volume (sum of reps x kg) per ISO week or month split by movement pattern, over a date range (default last 12 weeks).

get_workout

ChatGPT
One workout in full: stats, computed metrics (TRIMP, zone seconds, efficiency factor, aerobic decoupling, HR profile used), HealthKit metadata, and optionally downsampled 1 Hz series (kinds: hr, speed_mps, power_w, cadence, stride_m, alt_m).

get_workout_notes

ChatGPT
Notes attached to a workout (by the athlete or the assistant).

list_exercises

ChatGPT
The exercise catalogue (names, aliases, movement pattern) with the athlete's history per exercise: last trained day, current 5RM in kg, session count. Use it to pick the canonical name before log_performed.

list_generated_workouts

ChatGPT
Recent AI-generated workout option sets and which option the athlete chose.

list_workouts

ChatGPT
Workouts newest first with distance (m), duration (s), avg/max HR, load (TRIMP) and seconds in HR zones 1-5. Page with before=<start ISO timestamp from next_before>; filter with since and activity_type (e.g. running, functional_strength, walking).

log_performed

ChatGPT
Record what the athlete performed, on their behalf (logged_by 'mcp'). kind is 'strength', 'run' or 'cross'. A strength structure is {blocks: [...]}: a block with 'rounds' is a circuit done that many times through ("4x through" = rounds 4, each exercise listed once with its reps or seconds and load); a block without rounds holds straight sets, each exercise carrying 'sets' and 'reps' (or 'seconds'). Weights may be given in pounds or kilograms: load is {kg: 60} or {lb: 135} (converted to kg), or {percent_1rm}, {rpe}, {bodyweight: true}. Example: {"blocks": [{"exercises": [{"name": "Bench press", "sets": 5, "reps": 5, "load": {"lb": 135}}]}, {"rounds": 4, "exercises": [{"name": "kettlebell swing", "reps": 15, "load": {"kg": 24}}, {"name": "ski pull", "seconds": 45}]}]}. Use catalogue names (check with resolve_exercises; equipment words like barbell/dumbbell/kb are ignored when matching). Unknown names are rejected with suggestions unless create_missing is true, in which case the API infers the movement from the name or uses the 'movement' you give (push, pull, squat, hinge, carry, core, cardio) and lists them in created_exercises; prefer create_exercise for anything genuinely new. A name already in the catalogue as written always binds to that row. Runs and cross use {steps: [...]}: each step has a length (minutes or seconds, and/or a distance as metres, km or mi; a treadmill run is {minutes: 35, mi: 3.4}) and a target ({zone}|{hr:{lo,hi}}|{pace:{lo,hi} s/km}|{power:{lo,hi}}|{rpe}), or {repeat: n, steps: [...]}. date is YYYY-MM-DD (default: the workout's day, else today in tz); the sets attach to that day's strength workout when there is one.

metric_series

ChatGPT
A daily metric over a date range (YYYY-MM-DD), rolled up by day, week or month. Metrics: readiness_v2 (0-100, the sum of four inputs; readiness_v1 is the old formula), rhr_7d, rhr_7d_z, hrv_7d, hrv_7d_z, sleep_h, sleep_30d_mean, trimp_day, acwr, vo2max_latest, vo2max_slope_90d, weekly_km, weekly_min, weekly_trimp, weekly_climb_m, climb_m (metres of ascent), body_mass, race_pace_1609 / 3000 / 5000 / 10000 / 21097 (s per km). Default range: last 90 days.

propose_cycle

ChatGPT
Propose a 2–16 week training cycle (stored PROPOSED until the athlete approves it). start is a Monday, end a Sunday (YYYY-MM-DD). focus: aerobic_base, threshold, vo2, race_prep, return_to_run, maintenance, strength. weekly_template: one entry per week, {"week": 1, "minutes": 240, "trimp": 300, "sessions": 5, "deload": false, "km"?, "zone_share"?: {"z1".."z5" fractions}, "long_run_min"?}. rules: {"max_sessions_per_week"?, "rest_days"?: ["mon"], "acwr_ceiling": 1.3, "min_easy_share": 0.75}. success_criteria maps a linked goal set / goal id to the expected progress fraction at cycle end.

propose_goal

ChatGPT
Propose one measurable goal: a metric, a target value and a date. The baseline is taken from the athlete's data today. metric is one of vo2max_latest (increase), rhr_7d (decrease), hrv_7d (increase), sleep_h (increase), weekly_km / weekly_min (increase), weekly_climb_m (increase; metres of ascent per week), body_mass (either; pass direction), exercise_rm (increase; spec.exercise + spec.reps, kg), race_pace (decrease; spec.distance_m one of 1609, 3000, 5000, 10000, 21097; value is seconds per km, the fastest measured or Riegel-predicted effort in the window), efficiency_factor (accepted, not yet evaluated). horizon: near (weeks), season (months), year. Units are SI: vo2max_latest mL/kg/min, rhr_7d bpm, hrv_7d ms, sleep_h hours, weekly_km km per week, weekly_min minutes per week, body_mass kg, exercise_rm kg (spec.exercise + spec.reps). Dates are YYYY-MM-DD. The result is stored as a PROPOSAL (status 'proposed', inactive) until the athlete approves it in the app; nothing is activated or superseded.

propose_goal_set

ChatGPT
Propose an overarching aim ('Improve base cardio') as a goal set: a title, a description of why and what better means, an optional overall target_by, achieve_at (share of weight that must be achieved to call the set met, default 1.0), and components: a list of goals {title, metric, target_value, target_by, weight, horizon?, notes?, direction?, spec?}. Weights are relative within the set. metric is one of vo2max_latest (increase), rhr_7d (decrease), hrv_7d (increase), sleep_h (increase), weekly_km / weekly_min (increase), weekly_climb_m (increase; metres of ascent per week), body_mass (either; pass direction), exercise_rm (increase; spec.exercise + spec.reps, kg), race_pace (decrease; spec.distance_m one of 1609, 3000, 5000, 10000, 21097; value is seconds per km, the fastest measured or Riegel-predicted effort in the window), efficiency_factor (accepted, not yet evaluated). Units are SI: vo2max_latest mL/kg/min, rhr_7d bpm, hrv_7d ms, sleep_h hours, weekly_km km per week, weekly_min minutes per week, body_mass kg, exercise_rm kg (spec.exercise + spec.reps). Dates are YYYY-MM-DD. The result is stored as a PROPOSAL (status 'proposed', inactive) until the athlete approves it in the app; nothing is activated or superseded.

propose_workout

ChatGPT
Add one workout to the standalone schedule for a specific date outside any cycle. This writes the planned session immediately; use only for an explicitly requested schedule addition. day is YYYY-MM-DD (today ± a few days). session is one planned session without 'day': {"kind": easy|long|tempo|intervals|recovery|strength|cross, "title", "duration_min"?, "zone"?, "load_estimate_trimp"?, "notes"?, "structure"?} with the same structure shapes as save_weekly_plan (strength: blocks with rounds for circuits). Stored immediately as ACTIVE on the week's standalone plan (created or appended) and shown on the Today page and the Upcoming/Schedule views; no approval step. For several sessions or several days use schedule_sessions; for cycle weeks use save_weekly_plan or add_sessions_to_plan. Strength exercise names must resolve against the catalogue (resolve_exercises); unknown names are rejected with suggestions — create them first with create_exercise, deciding the movement yourself.

recompute_goals

ChatGPT
Re-evaluate every active goal and goal set right now from the latest daily metrics and logged strength sets, store today's progress point for each goal, and persist any achieved / missed transition. Returns the transitions plus the refreshed goals. The same step runs automatically each night and after every synced workout.

remove_plan_session

ChatGPT
Remove one session of a week, found by day (mon..sun) and title. New version; a standalone week left empty is abandoned. To move a session to another week, remove it here and schedule_sessions it on the new date.

render_coach

ChatGPT
Show Wolf's interactive coaching visual after get_coaching_context. Rereads authoritative existing data; does not write. view: overview/goals/cycle/session/progress/adjust/proposal. proposal displays saved proposed weeks for athlete review. Optional concise feedback from the conversation (max 600 chars); include human context and long-term connection, never fabricate evidence. User can explore weeks/workouts and expand Wolf targets. Changes are discussed in chat.

replace_performed_log

ChatGPT
Correct any performed log (run, strength or cross; with or without a workout or plan): replaces the log whose id you pass (log_performed's result, a workout's strength logs, or get_plan_compliance's session log_id) with a new structure in the same shape log_performed takes for that kind (blocks for strength, steps for runs). The old log and its sets are voided, never deleted; day, workout, plan reference, kind and start/end times carry over; notes are replaced when given. Use when the athlete says a set, weight, distance or note was recorded wrong.

replace_plan_session

ChatGPT
Replace one session of a week, found by day (mon..sun) and title, with a new session spec (same shape as save_weekly_plan's sessions; give it a different day to move it within the week). New version; status unchanged, so active weeks update live.

resolve_exercises

ChatGPT
Check exercise names against the catalogue before writing a plan or a log. For each name: the catalogue entry it maps to and how (exact, stripped of equipment words, or a close spelling), or null with suggestions and the movement the API would infer for a new entry. Read-only. Use the returned canonical names in save_weekly_plan / propose_workout / log_performed; create genuinely new exercises with create_exercise first.

review_weekly_proposal

ChatGPT
Apply an explicit athlete approval or keep-current action from the Wolf visual. Rejects stale reviews; never changes goals or cycle status.

save_weekly_plan

ChatGPT
Save a structured weekly plan (created_by mcp). week_start is a Monday. sessions: [{"day": "mon".."sun", "kind": easy|long|tempo|intervals|recovery|strength|cross|rest, "title", "duration_min"?, "distance_m"?, "zone"? 1-5, "load_estimate_trimp"?, "notes"?, "structure"?}]. Run/cross structure: {"steps": [step]} where a step is {"name"?, one of minutes|seconds|metres|until:{"hr_below"}, "target": one of {"zone"}|{"hr":{"lo","hi"}}|{"pace":{"lo","hi"} s/km}|{"power":{"lo","hi"}}|{"rpe"}} or {"repeat": n, "steps": [step]}. Strength structure: {"blocks": [{"rounds"?, "rest_s"?, "exercises": [{"name", "sets"?, "reps"?, "seconds"?, "load"?: {"kg"}|{"percent_1rm"}|{"rpe"}|{"bodyweight": true}, "rest_s"?}]}]}. Totals are derived from steps; a duration or distance that disagrees by more than 10 % is rejected. With cycle_id the week gets week_no and targets from the cycle template and is PROPOSED until approved. WITHOUT cycle_id the sessions are merged into the week's existing standalone plan (same day + title skipped) and are active at once, so sending the whole week again never duplicates a day; prefer schedule_sessions or add_sessions_to_plan to add to a week. targets: {minutes, km?, trimp, zone_share?, sessions}. Strength exercise names must resolve against the catalogue (resolve_exercises; equipment words are ignored); unknown names are rejected with suggestions — create them first with create_exercise, deciding the movement.

schedule_sessions

ChatGPT
Schedule any number of dated sessions outside a cycle, e.g. a run tomorrow and a strength session on Thursday. items: [{"date": YYYY-MM-DD, "session": {same shape as propose_workout's session, without day}}]. Dates from 2 days ago to 28 days ahead. Each date is filed into its week's standalone plan (created if needed, otherwise a new version), stored ACTIVE at once: no plan setup, no approval. Check get_schedule first so you do not double-book a day.

summary_today

ChatGPT
Today's picture: latest recovery and load metrics with the day each comes from, this week's volume, the last 7 days of load, and the most recent workout. tz is the athlete's IANA time zone (defaults to the server's).

void_performed_log

ChatGPT
Retire a performed log that should not exist: a misdated duplicate, a session logged twice. The log and its sets are marked voided (never deleted) and stop counting everywhere. To fix a log's contents use replace_performed_log instead.

volume_by_type

ChatGPT
Minutes, km and count per activity type per ISO week (or month) over a date range: training composition.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/wolf"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/wolf)