automations_list
ChatGPTList the scheduled automations the user set up for Titan (weekly_review, monthly_checkin, week_ahead), each with its assistant and schedule. Read it before setting automations up, so none is created twice.
automations_update
ChatGPTRecord which Titan automations now exist as scheduled tasks, after you created them (ChatGPT) or the user confirmed they created them (Claude). Send the complete list: it replaces the previous one; an empty list records none. schedule is a short human description such as "Sun 19:00" or "1st of the month 09:00". Titan then skips its own Sunday recap when a weekly_review exists, and its check-in reminder when a monthly_checkin exists. Repeating a call with the same mutationId and payload returns the original result.
body_weight_create
ChatGPTBody weight create in the connected Titan account. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Weight is in kilograms (20–500).
body_weight_delete
ChatGPTBody weight delete in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Weight is in kilograms (20–500). Requires explicit data:delete consent.
body_weight_entry
ChatGPTRead one owned weigh-in and its revision before correcting it. Deleted records return a tombstone.
body_weight_history
ChatGPTRead individual weigh-ins, not daily averages. Follow nextAfter until null, then sort measurements by date; UUID pagination is not chronological.
body_weight_target
ChatGPTRead the body-weight target in kg. Null means no target record; a null targetKg explicitly clears a target.
body_weight_target_update
ChatGPTSet a body-weight target in kg using its current revision (zero if absent). Null clears it. Reuse mutationId and exact payload for retries.
body_weight_update
ChatGPTBody weight update in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Weight is in kilograms (20–500).
checkin_decision_save
ChatGPTSave the outcome of a check-in (refused with not_enough_data while Overview has too little data to propose from; a new pending proposal supersedes an older one): status accepted or declined after the user explicitly said yes or no in this chat (apply an accepted change with its own tool first: Titan refuses accepted with change_not_applied until each change is in place; never accepted or declined in an unattended scheduled run), or pending when they want to think about it or the chat ends without an answer; the app then shows the proposal with Accept and Decline, and Accept applies it. summary follows the mandatory structure (goal, trajectory, whatWorks, whatBlocks, priority in one short line, proposedChange). changes list each change with before and after: kind calories, protein, carbs or fat (daily targets), weekly_target (sessions a week), weekly_sets (the weekly sets target as "low-high", e.g. "12-18", with muscle for one muscle’s override; before is the range it uses now, from coaching_profile: the muscle’s override, else the overall target, else "10-20"), weight_goal_kg, weight_goal_date (YYYY-MM-DD), or other (text, for example a routine change; never applied by the app, so not allowed in a pending proposal). Take before values from Titan, never from memory. Takes a fresh decisionId UUID. Repeating a call with the same mutationId and payload returns the original result.
checkin_history
ChatGPTList the user’s check-in decisions, newest first: date, assistant, summary, changes with before and after, status (pending, accepted or declined) and where it was validated (chat or app). Read it at the start of a check-in to see the last priority, and to learn whether a pending proposal was accepted.
coaching_profile
ChatGPTRead the user’s coaching profile in Titan ("My coaching"): goals (every goal they have; goal is the first), level and training history, days per week (the weekly session target) and minutes per session, equipment, nutrition tags and notes, and limitations (injuries, pain, exercises to avoid; null unless they consented). revision 0 means My coaching was never saved: goal, minutes per session and equipment then show the user’s setup answers from the app. Read it before a check-in or before planning, and never ask for what it already holds. weeklySets is the weekly sets target the Progress → Muscles screen draws (direct sets per muscle per week, with per-muscle overrides and who set it); null means Titan’s default, defaultWeeklySets (10-20).
coaching_profile_update
ChatGPTSave the coaching profile, only after the user agreed to what you will save. Takes the current revision (0 if none) and the fields to change: omitted fields keep their value, null clears one. goals lists every goal the user has (one to three; fat_loss, muscle and strength combine, maintenance stands alone) and sets goal to its first; send goals rather than goal when they have more than one. daysPerWeek is the weekly session target. limitations are health data: Titan refuses them with limitations_consent_required unless limitationsConsent is true, which you send only after asking the user, in these words, "Do you agree that Titan stores this health information to tailor your coaching?" and getting a clear yes. limitationsConsent false withdraws consent and deletes the limitations. weeklySets sets the weekly sets target shown on Progress → Muscles (whole sets per muscle per week, 0-40, low below high, optional per-muscle overrides; send the complete target, null returns to Titan’s 10-20); titan_guide topic checkin says when to change it. Repeating a call with the same mutationId and payload returns the original result.
exercise_archive
ChatGPTExercise archive in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
exercise_art_search
ChatGPTFind the exact picture key for exercises before creating or renaming them (exercise_create and exercise_update take artKey; it also picks the exercise’s demo clip). One call for every exercise of a routine: each query, one exercise name in any language ("Incline Smith press", "développé couché haltères"), returns the closest supported exercises, best first, with key, label, hint, kind (exercise, or category: a category demo standing in for any exercise of that kind) and clip (true when its mascot demo is live). sure means Titan reads that name with certainty and the first match is it. Pass the match that is exactly the movement; prefer one with clip when several are equally right; when none is, pass the category rather than a different lift.
exercise_create
ChatGPTExercise create in the connected Titan account. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
exercise_detail
ChatGPTRead one owned exercise.
exercise_update
ChatGPTExercise update in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
food_art_search
ChatGPTFind the exact picture keys for foods before logging them (meal_log, meal_add_items, nutrition_create, nutrition_update, foods_save, saved_meal_save take artKeys). One call for every food of a meal: each query, in any language ("pilons de poulet", "boiled potatoes"), returns the closest pictures, best first, with key, label, hint and kind (food, family: a general picture such as cat_poultry for any poultry, drink, condiment, generic). Pick the most exact; a family when nothing specific fits. Never pick a drink, sauce or flavour for the main food of a plate.
food_barcode
ChatGPTLook up a packaged food by its barcode digits (EAN-8, EAN-13, UPC-A or UPC-E). Returns the user’s saved food for that barcode if there is one (source "saved"), otherwise public product data from Open Food Facts (source "openfoodfacts", ODbL: credit Open Food Facts when you use it) or source "not_found". Only the barcode is sent to Open Food Facts. Product values are per serving and per 100 g; null is unknown.
foods_delete
ChatGPTDelete a saved food from the user’s "My foods", only when they ask. Takes the food’s current revision (foods_search). Meals already logged from it keep their values. Requires the data:delete permission. Repeating a call with the same mutationId and payload returns the original result.
foods_save
ChatGPTSave a food to the user’s "My foods", or correct one. A new food takes a fresh foodId UUID and expectedRevision 0; a correction takes its current revision (foods_search) and the complete food, and keeps its foodId. Give values per serving and/or per 100 g; omitted macros are unknown, not zero. Offer to save a food the user logs again or calls their usual, or a branded product you identified from a photo or description, and save only after they agree. When the user corrects a saved food (for example "my protein shake has 32 g protein"), update it. Repeating a call with the same mutationId and payload returns the original result.
foods_search
ChatGPTSearch the user’s saved foods ("My foods" in the Titan app) by name, brand or barcode, most recently used first (order "frequent" for most logged, "name" for A–Z); no query lists them. Call this first, before estimating, whenever the user mentions a food they may have saved, a brand or a packaged product. Each food has perServing and per100 values (per 100 g, or 100 ml when unit is ml) and servingSize; null is unknown, not zero. Scale them to the amount eaten and log with nutrition_create, passing the food’s id as foodId and a description like "Skyr natural · Fixture Dairy, 150 g". Follow nextOffset for more.
health_sleep
ChatGPTRead the user’s nights and recovery from Apple Health or Health Connect, as imported by the Titan app: per day, bedtime and wake time (local), minutes asleep and in bed, sleep stages (deep, core or light, REM, awake) when the device records them, resting heart rate, heart-rate variability (ms; SDNN on iPhone, RMSSD on Android) and steps. Read-only. Returns shared:false and no data unless the user turned on "Share with your AI" in Titan. Context for recovery and training load, not medical advice.
health_workouts
ChatGPTRead workouts the user recorded outside Titan (a run on their watch, a bike ride, a class), from Apple Health or Health Connect as imported by the Titan app: date, local start time, activity, duration, kcal, average and maximum heart rate, distance in meters and the app that recorded it. Also titanWorkouts: the heart rate during workouts logged in Titan, read from the watch for their time (average, maximum and heartRatePerMinute as [minute from start, bpm]); their sets are in workout_detail. Read-only. Returns shared:false and no data unless the user turned on "Share with your AI" in Titan.
meal_add_items
ChatGPTAdd items to a meal already logged, only when the user says so ("add a coffee to my lunch"). Find the meal and its mealId with nutrition_day. Each item takes a fresh entryId UUID and the same fields as in meal_log. Repeating a call with the same mutationId and payload returns the original result.
meal_card_refresh
ChatGPTRefresh the authoritative meal and day snapshot after a concurrent edit.
meal_card_update
ChatGPTUpdate the meal being edited in the Titan card. Requires its current revision and complete entry. Retry with the same mutationId and payload after uncertain delivery.
meal_delete
ChatGPTDelete a whole meal and all its items, only when the user asks. Takes the meal’s current revision (nutrition_day). Requires the data:delete permission. Repeating a call with the same mutationId and payload returns the original result.
meal_log
ChatGPTLog one meal: everything the user ate in one sitting, as one call with all its items (for example "200 g chicken, 100 g rice, a tbsp of andalouse" is one meal of 3 items). Every food the user names is its own item with its own amount and macros, matched to a saved food with foods_search when possible: "3 eggs, 2 slices of protein bread, 2 olive-oil sprays and a swirl of Sriracha" is 4 items, never one item listing them. One message about one sitting is one meal_log call; a message about several sittings ("eggs for breakfast and a salad for lunch") is one call per sitting. A lone item mentioned later is its own meal, unless the user says to add it to an existing meal (then meal_add_items). name: a short name from the main foods, for example "Chicken and rice". slot: mealSlotText eatenAt: HH:MM only when the user gave a time; never invent one. Each item takes a fresh entryId UUID, a description, calories and always protein, carbs and fat in grams, icon, and foodId when it comes from a saved food (foods_search). All items are saved together or none is. Takes a fresh mealId UUID. Repeating a call with the same mutationId and payload returns the original result.
meal_update
ChatGPTMove a meal to another slot, rename it or set when it was eaten, when the user asks. Takes the meal’s current revision (nutrition_day) and only the fields to change. slot: mealSlotText A slot the user chose in the Titan app always wins: Titan keeps it, and the result says slotKept; tell the user it stays where they put it. Repeating a call with the same mutationId and payload returns the original result.
nutrition_create
ChatGPTNutrition create in the connected Titan account. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Omitted macros mean unknown, not zero. A successful save includes its Titan meal card; do not call show_meal_card again or duplicate the card as a table.
nutrition_day
ChatGPTRead a nutrition day and totals. Follow nextAfter for more entries; use the user’s intended YYYY-MM-DD date.
nutrition_delete
ChatGPTNutrition delete in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Omitted macros mean unknown, not zero. Requires explicit data:delete consent.
nutrition_entry
ChatGPTRead one owned meal including its current revision before correcting it.
nutrition_history_search
ChatGPTFind foods the user logged before, in one call: for "same as last time", "the rice cakes I had" or "what I usually eat for breakfast". Searches their own logged items (case and accents ignored, singular or plural) from the last 180 days unless from/to (YYYY-MM-DD, at most two years apart) say otherwise, and returns each distinct food once, most recently eaten first, then most often: its name, last description (with the amount), foodId and icon, the last calories, protein, carbs and fat, lastEatenDate, lastSlot and timesLogged. Use it once, then log with meal_log reusing those values (scaled to the amount now eaten) and its foodId; never read nutrition days one by one to find a food. truncated means only the newest items were searched, from searchedFrom: narrow from/to for older ones.
nutrition_targets
ChatGPTRead daily nutrition targets. Null means no targets are configured.
nutrition_targets_update
ChatGPTSet all daily targets using the current revision (zero if absent). Null clears a target. Reuse mutationId and payload for retries.
nutrition_update
ChatGPTNutrition update in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Omitted macros mean unknown, not zero. A successful save includes its Titan meal card; do not call show_meal_card again or duplicate the card as a table.
progress_summary
ChatGPTRead the user’s check-in summary for the last 7 and 28 days: weight trend (7-day average now against 28 days ago, start weight and goal), sessions finished against their weekly target, top-set progress on their most-trained lifts, days with meals logged, data coverage, and a status (on_track, watch or not_enough_data) with the numbers behind it. Measured data only. Call it first for a check-in, progress review or "is it working?"; titan_guide topic checkin explains the check-in.
routine_archive
ChatGPTRoutine archive in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Requires explicit data:delete consent.
routine_create
ChatGPTRoutine create in the connected Titan account. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
routine_detail
ChatGPTRead one owned routine and its ordered exercises.
routine_update
ChatGPTRoutine update in the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
saved_meal_delete
ChatGPTDelete a saved meal, only when the user asks. Takes its current revision (saved_meals_list). Meals already logged from it keep their items. Requires the data:delete permission. Repeating a call with the same mutationId and payload returns the original result.
saved_meal_log
ChatGPTLog a saved meal as one meal with all its items, when the user says they ate it ("I had my usual breakfast", "lunch was lunch A"). Name it by savedMealId (saved_meals_list) or by name; a name that matches several is refused, so ask which one. Items from a saved food are worked out again from that food. portion scales every item (0.25 to 4, default 1). slot and eatenAt follow the same rule as meal_log: only from the user’s own words. Takes a fresh mealId and mutationId; repeating a call with the same mutationId returns the original result. To change one item afterwards, use nutrition_update on it.
saved_meal_save
ChatGPTSave a meal the user wants to log again, or change one, only after they agree. A new saved meal takes a fresh savedMealId and expectedRevision 0; a change takes its current revision (saved_meals_list) and the complete meal. To save a meal already logged, read its items with nutrition_day and pass them with their foodId. Give each item its values; with a foodId, also its amount. Repeating a call with the same mutationId and payload returns the original result.
saved_meals_list
ChatGPTList the user’s saved meals ("Meals" in My foods in the Titan app), most used lately first, optionally searched by name. Each has its name, items (description, values, and the saved food and amount it comes from), totals and revision. Call it when the user names a meal they may have saved ("my usual breakfast", "lunch A").
set_create
ChatGPTLog one completed set (weight and reps) for an exercise in a Titan workout. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Supply the current workoutRevision.
set_delete
ChatGPTDelete one logged set from a Titan workout. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Supply the current workoutRevision. Requires explicit data:delete consent.
set_update
ChatGPTCorrect the weight, reps or RPE of a set already logged in a Titan workout. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Supply the current workoutRevision.
show_meal_card
ChatGPTShow an existing saved meal with daily calories/macros and an inline editor when the user asks to view it. Meal create/update already includes this card: do not render a duplicate after those saves. Fetches authoritative account data; never pass invented display values.
show_weight_progress
ChatGPTDisplay a weight chart for the user’s intended date interval (YYYY-MM-DD), measured coverage and change. No extrapolation or invented measurements.
show_workout_card
ChatGPTDisplay the next routine or an explicitly requested routine with its saved exercises, sets, rep ranges, rest and cues.
titan_guide
ChatGPTExplains what Titan can do and how to use its tools: training, meals and nutrition targets, saved foods and barcodes, body weight and goal, and cards, with the rules for saving changes. Call it first when the user asks what Titan can do, or before using a Titan feature you are unsure about. Optional topic narrows it to one area.
training_overview
ChatGPTFind the connected user’s saved workout routines by name, next rotation workout and active session. Use before starting a named routine (for example "Start Reviewer Upper"), planning training or answering "what is my next workout?". Match the saved name to its ID; ask if multiple routines match.
weekly_target_update
ChatGPTSet how many sessions a week the user plans (1-7), or null to unset it, only after they agree. Takes training_overview’s profileRevision as expectedRevision. Progress compares finished sessions with it. Repeating a call with the same mutationId and payload returns the original result.
workout_cancel
ChatGPTCancel the connected account’s active workout without saving it as completed. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Requires explicit data:delete consent.
workout_delete
ChatGPTDelete a finished or cancelled workout from the connected Titan account. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
workout_detail
ChatGPTRead one owned workout, its set results and prescriptions.
workout_effort
ChatGPTSet, change or clear how hard a finished workout felt (effort 1 easy to 5 brutal), only when the user tells you; the Titan app asks after each workout and the user can change it there too. Takes the record’s current expectedRevision; updates replace the complete record definition. Values are the measurements the user reports. Repeating a call with the same mutationId and payload returns the original result.
workout_exercise_remove
ChatGPTRemove an exercise from the connected account’s active workout. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Requires explicit data:delete consent.
workout_exercise_set
ChatGPTAdd an exercise to the connected account’s active workout, or change that exercise’s position, sets, reps or rest. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
workout_finish
ChatGPTFinish the connected account’s active workout. Read the current record first and supply its current expectedRevision. Updates require the complete record definition. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries.
workout_history
ChatGPTRead a bounded page of workout history. Follow the returned cursor for older sessions.
workout_start
ChatGPTStart a workout in the connected Titan account from one of its saved routines. Use a new entity UUID and expectedRevision 0. Use the user’s intended date and measured values; never invent missing measurements. Reuse mutationId and the exact payload on retries. Start a saved workout routine when the user says "start [routine name]" or "begin my workout". Resolve its ID using training_overview; do not invent a routine or start duplicate active sessions.
automations_list
Claudeautomations_update
Claudebody_weight_create
Claudebody_weight_delete
Claudebody_weight_entry
Claudebody_weight_history
Claudebody_weight_target
Claudebody_weight_target_update
Claudebody_weight_update
Claudecheckin_decision_save
Claudecheckin_history
Claudecoaching_profile
Claudecoaching_profile_update
Claudeexercise_archive
Claudeexercise_art_search
Claudeexercise_create
Claudeexercise_detail
Claudeexercise_update
Claudefood_barcode
Claudefoods_delete
Claudefoods_save
Claudefoods_search
Claudemeal_add_items
Claudemeal_card_refresh
Claudemeal_card_update
Claudemeal_delete
Claudemeal_log
Claudemeal_update
Claudenutrition_create
Claudenutrition_day
Claudenutrition_delete
Claudenutrition_entry
Claudenutrition_history_search
Claudenutrition_targets
Claudenutrition_targets_update
Claudenutrition_update
Claudeprogress_summary
Clauderoutine_archive
Clauderoutine_create
Clauderoutine_detail
Clauderoutine_update
Claudesaved_meal_delete
Claudesaved_meal_log
Claudesaved_meal_save
Claudesaved_meals_list
Claudeset_create
Claudeset_delete
Claudeset_update
Claudeshow_meal_card
Claudeshow_weight_progress
Claudeshow_workout_card
Claudetitan_guide
Claudetraining_overview
Claudeweekly_target_update
Claudeworkout_cancel
Claudeworkout_delete
Claudeworkout_detail
Claudeworkout_exercise_remove
Claudeworkout_exercise_set
Claudeworkout_finish
Claudeworkout_history
Claudeworkout_start
Claude