Health

Apple Health only: list every synced-health metric type supported by Health. This is a capability catalog, may include metrics for which the user has no data, and does not search provider medical records or user-reported reconciliation data. Prefer `view_synced_metrics` when the question depends on what is actually available for this user.

view_all_metrics

Get whether the authenticated user currently has an active synced activity data link.

get_apple_health_connector_status

Initiate the OAuth flow for a health data provider.

initiate_organization_oauth

Delete a single Health account connection by connection ID.

delete_account_connection

Generate insights from the authenticated user's available Health resources.

generate_insights

Load configuration for the Health app.

config

Apple Health only: aggregate synced quantity metrics over a specific time window. Use for de-duplicated totals, exact counts, custom time buckets, and ordered trends. This uses server-side aggregation; do not substitute a sum of overlapping raw samples. For summed metrics with multiple sources, prefer the returned `multi-source-merge` rows when present. Check result_status and its completeness and row-limit signals before treating a result as complete. Use returned local timestamps as-is; canonical *_ms timestamps use UTC Unix milliseconds. Small results remain inline. Large results include a summary and an authorized JSON file reference. Use the file location provided by the current runtime; do not assume a fixed filesystem path. This does not search provider medical records or user-reported reconciliation data. Use query_samples_file for raw or category samples.

aggregate_timeseries_file

Apple Health only: return raw synced samples for one or more exact metrics over a specific time window. Small results remain inline. Large results use a short summary and an authorized JSON file reference that the current runtime can make available for analysis. This does not search provider medical records or user-reported reconciliation data. Use after `health_summary` for sample-level analysis and category metrics such as sleep and workouts, unless the request is already limited to an exact Apple Health metric and date window; in that case, call this action directly. For a trend, pattern, or what-stands-out request, compare an appropriate recent period with older available data in separate, non-overlapping windows rather than stopping at a recent-only slice or overlapping summary aggregates. Use `view_synced_metrics` date coverage to choose those windows, and retrieve actual samples because coverage counts are not trend evidence. Use exact synced metric identifiers such as `sleepAnalysis`, `workout`, or `heartRate`. When `time_local`, `start_time_local`, or `end_time_local` is present, use it as-is for display; it already includes the effective fixed timezone offset. When `timestamp_semantics` is `utc_unix_milliseconds`, the corresponding `*_ms` fields are canonical UTC instants. For legacy rows with neither local fields nor that marker, treat `*_ms` as already-localized wall-clock values. Never apply an additional offset to a local field or legacy timestamp. For `sleepAnalysis`, group adjacent asleep segments into sleep sessions, starting a new session when the gap exceeds about 90 minutes, and attribute each session to the local date of its end time. If raw samples show multiple sources and the answer needs a de-duplicated total or trend, follow up with `aggregate_timeseries_file`. Inspect the attached data before answering rather than relying only on inline summary rows. Use the runtime-provided file location when one is returned; do not assume a fixed filesystem path.

query_samples_file

Search for providers to connect medical records in Health.

search_providers

Search RxNorm ingredient concepts for current-medication reconciliation.

search_medications