get_athlete_insights
ClaudeSUPPLEMENTARY tool. Returns 3-5 emoji-prefixed AI coaching bullets for one athlete-race in two flavors via the type arg: - segment: per-segment performance vs the field (rankings, percentile patterns, terrain-aware strengths/weaknesses). - passing: per-segment passed / passed-by analysis correlated with terrain (hill racing, late-race execution). IMPORTANT — this is NOT a substitute for get_athlete_result. The athlete-result MCP App ALREADY shows these insights inline in the rendered card, so explicitly calling this tool from the chat is rarely necessary. Only call it when (a) the user explicitly asks for additional coaching analysis beyond what the card shows, OR (b) you've already called get_athlete_result and need extra qualitative material to enrich your prose. Do NOT call this as a shortcut to skip get_athlete_result — that bypasses the interactive result card entirely. Returns plain text with bullets separated by blank lines, each starting with an emoji. The same MyRace-incomplete-dataset and distance-comparability caveats from search_athletes apply.
get_athlete_result
Claudeget_elevation_profile
Claudeget_race_info
Claudeget_race_results
Claudelist_curated_race_issues
ClaudeREAD-ONLY curation-maintenance worklist from MyRace's curated race calendar. Returns two sections: missingEvent — active curated races with no event linked (each should have its event find-or-created and linked); and duplicates — groups of active curated races sharing the same natural key (name + date + distance) where more than one active row exists, each member annotated with publish_status, event_id, slug, completeness, and created_at so you can apply the keeper rule (published > draft > has event+slug > highest completeness > oldest) and deactivate the rest. Only exact-key duplicates are surfaced; use search_curated_races for fuzzier matches. Call this to drive the on-demand cleanup mode.
list_race_distances
ClaudeList MyRace's canonical race distances (name + meters). Call this before proposing a curated race so you use a distanceName + distanceMeters that match a registered distance (this is how curated races map to the distance dropdown). If a race's distance isn't listed, propose it anyway — the write path adds it as a custom distance.
list_races
ClaudeList every race in the MyRace database with race IDs, names, dates, distances, locations, and finisher counts. Call this first when the user asks about available races, or whenever you need to discover a valid race ID before using another tool. Results are ordered by race date, MOST RECENT FIRST. When the user names a race/distance WITHOUT a year (e.g. 'broken arrow 23k'), default to the most recent matching edition — the FIRST entry matching that race name and distance — and state which edition you used (e.g. 'Showing the 2026 edition, the most recent.'). Do NOT ask the user to pick a year, and do NOT assume an older year is the latest. Match on race name AND distance, since one event has many distances per year. Only use an older edition if the user explicitly asks for a specific year. IMPORTANT: This list is NOT exhaustive of every race in the world — it is only the set of races MyRace has currently imported. If a user asks about a race that isn't in the result, say so plainly and suggest related races from the list rather than guessing or fabricating data.
search_athletes
ClaudeLook up athletes by name to discover their raceId + pid pair. Use this as a NAME-TO-ID LOOKUP STEP — the response includes basic info (race name, date, finish time, overall + gender place) but is intentionally lightweight. IMPORTANT — for any 'how did X do at Y?' question, ALWAYS follow up with get_athlete_result using the pid and raceId from this response. get_athlete_result returns the comprehensive breakdown (splits, pacing analysis, ranking progression, qualifying status, etc.) AND triggers the interactive result card the user sees inline. Skipping get_athlete_result means the user sees only prose with no rich result card. Do NOT compose a full race-recap response from search_athletes data alone — call get_athlete_result first. For every time field, the response includes both a formatted string (e.g. finishTime: '3:05:35') for display AND a raw-seconds number (finishTimeSeconds: 11135) for math. Always use the *Seconds field when comparing, sorting, averaging, or computing differentials — never parse the formatted string yourself. IMPORTANT — comparing times across races: 1. MyRace's database is incomplete: it does NOT contain every race an athlete has ever run. It only contains the races MyRace has imported. Use list_races to see the full set of races covered. 2. Each result includes raceDistance (e.g., 'Marathon', 'Half Marathon', '10K', '50K'). Finish times are ONLY comparable between races of the same distance. 3. Because of (1) and (2), do NOT claim an athlete has a 'PR' or 'personal record' based on these results. The fastest time you see is the fastest time within the MyRace data — not necessarily the athlete's true PR. Phrase claims as 'her fastest marathon in MyRace' or 'her best CIM time' or 'among the races MyRace has data for, her fastest marathon was…' — never just 'PR' or 'personal record'. 4. When grouping an athlete's results, always group by raceDistance first; never compare a marathon time to a half marathon, 10K, or trail-race time.
search_curated_races
ClaudeSearch MyRace's CURATED race calendar (upcoming curated races shown on myrace.ai/tools/race-calendar) by name, location, distance, or date range. Use this to check whether a race is ALREADY in MyRace before proposing it, and to read an existing curated race's current field values. Each result includes its id (use as existingCuratedRaceId when proposing an augmentation) and a missingFields list showing which key fields are empty on that row. This is separate from list_races/get_race_info, which cover races MyRace has already imported RESULTS for.