# Delx Wellness Full Agent Context

Open-source MCP connectors for WHOOP, Oura, Garmin, Strava, Fitbit, Google Health, Withings, Apple Health, Samsung Health and Polar, plus Nourish nutrition context, hosted cloud beta and one-command profile packs for Hermes and OpenClaw. Local-first remains free. Hosted cloud adds OAuth, remote MCP, persistence, self-service MCP keys and audit controls for users who do not want to operate the stack.

## Project

- Repository: https://github.com/davidmosiah/delx-wellness
- License: https://github.com/davidmosiah/delx-wellness/blob/main/LICENSE
- Canonical domain: https://wellness.delx.ai
- Architecture: independent local-first MCP connectors and wellness modules, each installed from its own package or repo.
- Privacy stance: provider tokens stay on the user's machine or server; agents receive MCP responses, not OAuth secrets.
- Hosted beta: optional paid cloud layer for OAuth, remote MCP, persistence, self-service MCP keys, audit controls and support. Local open-source connectors remain free.
- Hosted hub staging: https://api.delx.ai/wellness-hub with MCP endpoint at https://api.delx.ai/wellness-hub/mcp.
- Trust pages: https://wellness.delx.ai/security and https://wellness.delx.ai/docs/cloud.

## Connector catalog

### WHOOP

- Package: whoop-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0
- Summary: A local-first MCP connector that lets agents reason over WHOOP recovery, sleep, cycles and workouts without handing OAuth tokens to the model.
- Best for: Best for recovery-led athletes, coaches and personal agents that need a daily readiness, sleep and load briefing before recommending training.
- Signals: Recovery, HRV, Sleep, Strain, Workouts
- Install: npx -y whoop-mcp-unofficial setup
- Doctor: npx -y whoop-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/whoop | GitHub: https://github.com/davidmosiah/whoop-mcp | npm: https://www.npmjs.com/package/whoop-mcp-unofficial | Docs: https://whoopmcp.vercel.app/ | Provider docs: https://developer.whoop.com/api/

Data availability:
- Recovery score, HRV, RHR, SpO2, skin temperature: available. Available when WHOOP returns a scored recovery.
- Sleep sessions, stages and performance: available. All scored sleep records.
- Cycles, day strain and kilojoules: available. Physiological cycle records.
- Workouts, sport and heart-rate zones: available. All recorded workouts exposed by the public API.
- Profile and body measurements: available. Height, weight and max heart rate.
- Continuous heart-rate or live BLE telemetry: not_available. Not exposed by the public WHOOP API; this connector is not a Bluetooth listener.

Agent tools:
- Profile and onboarding: `whoop_profile_get`, `whoop_profile_update`, `whoop_onboarding`
- Start here: `whoop_connection_status`, `whoop_daily_summary`, `whoop_weekly_summary`
- Auth and diagnostics: `whoop_capabilities`, `whoop_agent_manifest`, `whoop_privacy_audit`, `whoop_cache_status`, `whoop_get_auth_url`, `whoop_exchange_code`, `whoop_revoke_access`
- Profile: `whoop_get_profile`, `whoop_get_body_measurements`
- Collections: `whoop_list_recoveries`, `whoop_list_sleeps`, `whoop_list_cycles`, `whoop_list_workouts`
- Single records: `whoop_get_cycle`, `whoop_get_sleep`, `whoop_get_workout`, `whoop_get_cycle_sleep`, `whoop_get_cycle_recovery`

Privacy:
- Read-only by default.
- OAuth secrets and refresh tokens stay on the user's machine or server.
- Collection tools support summary, structured and raw response formats.
- The connector distinguishes upstream raw JSON from raw sensor telemetry.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- No live sensor stream.
- No raw accelerometer or Bluetooth data.
- Requires a WHOOP account and developer app.

### Oura Ring

- Package: oura-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0
- Summary: A privacy-aware Oura connector for agents that need sleep, recovery and daily activity signals from the Oura Cloud API.
- Best for: Best for sleep-first wellness agents, habit coaches and daily planning assistants that need readiness and overnight recovery context.
- Signals: Readiness, Sleep, Activity, HRV, SpO2
- Install: npx -y oura-mcp-unofficial setup
- Doctor: npx -y oura-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/oura | GitHub: https://github.com/davidmosiah/ouramcp | npm: https://www.npmjs.com/package/oura-mcp-unofficial | Docs: https://ouramcp.vercel.app/ | Provider docs: https://cloud.ouraring.com/docs/authentication

Data availability:
- Daily readiness score and contributors: available. Requires daily scope.
- Daily sleep score and sleep periods: available. Requires daily and/or sleep scope.
- Sleep stages and timing: available. Available when Oura returns scored sleep.
- Daily activity, steps, calories and MET: available. Requires daily scope.
- Heart-rate time series: conditional. Depends on ring, membership and granted scope.
- HRV and SpO2: available. HRV is surfaced through readiness; SpO2 requires spo2 scope and supported devices.
- Continuous sensor telemetry: not_available. Not exposed by the Oura Cloud API.

Agent tools:
- Profile and onboarding: `oura_profile_get`, `oura_profile_update`, `oura_onboarding`
- Start here: `oura_connection_status`, `oura_daily_summary`, `oura_weekly_summary`
- Auth and diagnostics: `oura_capabilities`, `oura_agent_manifest`, `oura_privacy_audit`, `oura_cache_status`, `oura_get_auth_url`, `oura_exchange_code`, `oura_revoke_access`
- Profile: `oura_get_personal_info`
- Daily collections: `oura_list_daily_readiness`, `oura_list_daily_sleep`, `oura_list_daily_activity`, `oura_list_daily_spo2`
- Detailed collections: `oura_list_sleep`, `oura_list_workouts`, `oura_list_heartrate`, `oura_list_sessions`, `oura_list_tags`

Privacy:
- OAuth tokens stay local.
- The connector is scoped to the permissions the user grants.
- Agent-facing summaries can avoid sensitive raw payloads.
- Raw means upstream Oura API JSON, not raw ring sensor streams.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- No continuous raw sensor telemetry.
- Availability depends on Oura device, membership and scopes.

### Garmin Connect

- Package: garmin-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Local Garmin Connect
- Summary: A Garmin Connect MCP server for agents that need rich wellness, sleep, stress, Body Battery, training and activity context.
- Best for: Best for endurance athletes and daily agents that need a broad view of sleep, stress, Body Battery, training readiness and recent activities.
- Signals: Sleep, HRV, Body Battery, Stress, Training Readiness
- Install: npx -y garmin-mcp-unofficial setup --auth
- Doctor: npx -y garmin-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/garmin | GitHub: https://github.com/davidmosiah/garminmcp | npm: https://www.npmjs.com/package/garmin-mcp-unofficial | Docs: https://garminconnectmcp.vercel.app/ | Provider docs: https://developer.garmin.com/gc-developer-program/health-api/

Data availability:
- Sleep duration, stages and score: available. Available when supported by the device/account.
- HRV status and overnight HRV: available. Available on supported devices/accounts.
- Body Battery and events: available. Charge and drain context for each day.
- Stress samples and daily summary: available. Per-day stress context.
- Training readiness and training status: conditional. Depends on device/account support.
- Activities, details, splits and zones: available. Recent activities and detail payloads.
- Continuous device telemetry: not_available. Not exposed by Garmin Connect web endpoints.

Agent tools:
- Profile and onboarding: `garmin_profile_get`, `garmin_profile_update`, `garmin_onboarding`
- Start here: `garmin_connection_status`, `garmin_daily_summary`, `garmin_weekly_summary`
- Auth and diagnostics: `garmin_capabilities`, `garmin_agent_manifest`, `garmin_auth_instructions`, `garmin_privacy_audit`
- Profile and devices: `garmin_get_profile`, `garmin_get_user_settings`, `garmin_list_devices`, `garmin_get_primary_training_device`
- Daily wellness: `garmin_get_daily_summary`, `garmin_get_steps_day`, `garmin_get_sleep_day`, `garmin_get_heart_day`, `garmin_get_hrv_day`, `garmin_get_stress_day`, `garmin_get_body_battery_day`, `garmin_get_training_readiness_day`
- Activities and body: `garmin_list_activities`, `garmin_get_activity_details`, `garmin_get_weight_range`

Privacy:
- No Garmin developer app is required.
- The MCP does not store the user's Garmin password.
- Local tokens are saved with user-only permissions.
- Private endpoint drift should be treated as integration drift, not user error.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Garmin can change private auth or endpoints.
- No accelerometer, gyroscope or raw device stream.

### Strava

- Package: strava-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0
- Summary: A Strava MCP connector that turns recorded activities and activity streams into agent-ready training context.
- Best for: Best for endurance agents, training reviews and coaches that need route-aware activity details without exposing GPS by default.
- Signals: Activities, Streams, Routes, Athlete Stats
- Install: npx -y strava-mcp-unofficial setup
- Doctor: npx -y strava-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/strava | GitHub: https://github.com/davidmosiah/strava-mcp | npm: https://www.npmjs.com/package/strava-mcp-unofficial | Docs: https://stravamcp.vercel.app/ | Provider docs: https://developers.strava.com/docs/reference/

Data availability:
- Activities: available. Runs, rides, swims, walks and workouts.
- Activity details, zones and splits: available. Heart rate, power, cadence, elevation and gear when present.
- Activity streams: available. Per-second samples for supported streams.
- GPS lat/lng streams: opt_in. Hidden by default; requires include_gps=true or raw mode.
- Athlete profile, zones and aggregate stats: available. Authenticated athlete context.
- Live device telemetry: not_available. Not exposed by Strava's public API.

Agent tools:
- Profile and onboarding: `strava_profile_get`, `strava_profile_update`, `strava_onboarding`
- Start here: `strava_connection_status`, `strava_daily_summary`, `strava_weekly_summary`
- Auth and diagnostics: `strava_capabilities`, `strava_agent_manifest`, `strava_privacy_audit`, `strava_cache_status`, `strava_get_auth_url`, `strava_exchange_code`, `strava_revoke_access`
- Athlete and training: `strava_get_athlete`, `strava_get_zones`, `strava_get_athlete_stats`
- Activities and streams: `strava_list_activities`, `strava_get_activity`, `strava_get_activity_zones`, `strava_get_activity_streams`
- Routes and context: `strava_list_routes`, `strava_get_route`, `strava_list_clubs`, `strava_get_gear`

Privacy:
- GPS streams are hidden unless explicitly requested.
- OAuth tokens stay local.
- Route geometry is redacted in summary and structured modes.
- Agents can start from daily or weekly summary tools before fetching details.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- No live telemetry.
- GPS requires explicit opt-in.
- API visibility depends on Strava scopes and privacy settings.

### Fitbit

- Package: fitbit-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0
- Summary: A Fitbit MCP connector for everyday health and activity agents that need sleep, heart, movement, body and nutrition context.
- Best for: Best for broad wellness agents that combine activity, sleep, HRV, breathing, weight and food/water logs.
- Signals: Activity, Sleep, Heart Rate, HRV, SpO2, Weight
- Install: npx -y fitbit-mcp-unofficial setup
- Doctor: npx -y fitbit-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/fitbit | GitHub: https://github.com/davidmosiah/fitbitmcp | npm: https://www.npmjs.com/package/fitbit-mcp-unofficial | Docs: https://fitbitmcp.vercel.app/ | Provider docs: https://dev.fitbit.com/build/reference/web-api/

Data availability:
- Daily activity: available. Steps, calories, distance and zones.
- Activity logs: available. Logged workouts.
- Sleep and sleep stages: available. When Fitbit returns stage data.
- Resting heart rate and daily zones: available. Available on scored days.
- Intraday heart-rate samples: conditional. Only when app/API access permits intraday data.
- HRV, SpO2 and breathing rate: conditional. Depends on supported device/account.
- Food, water, weight and body composition: available. When logged.
- Continuous device telemetry: not_available. Not exposed by Fitbit's public API.

Agent tools:
- Profile and onboarding: `fitbit_profile_get`, `fitbit_profile_update`, `fitbit_onboarding`
- Start here: `fitbit_connection_status`, `fitbit_daily_summary`, `fitbit_weekly_summary`
- Auth and diagnostics: `fitbit_capabilities`, `fitbit_agent_manifest`, `fitbit_privacy_audit`, `fitbit_cache_status`, `fitbit_get_auth_url`, `fitbit_exchange_code`, `fitbit_revoke_access`
- Profile and devices: `fitbit_get_profile`, `fitbit_list_devices`
- Activity and sleep: `fitbit_get_activity_day`, `fitbit_list_activities`, `fitbit_get_activity`, `fitbit_get_sleep_day`, `fitbit_list_sleep`
- Heart, body and nutrition: `fitbit_get_heart_day`, `fitbit_get_heart_intraday`, `fitbit_get_hrv_day`, `fitbit_get_spo2_day`, `fitbit_get_breathing_rate_day`, `fitbit_get_weight_day`, `fitbit_get_food_day`, `fitbit_get_water_day`

Privacy:
- OAuth tokens stay local.
- The connector only accesses granted Fitbit scopes.
- Intraday data is conditional and should be requested deliberately.
- Raw means upstream Fitbit JSON, not raw device sensor streams.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Intraday access can be limited.
- No continuous telemetry.
- Data depends on Fitbit device/account support.

### Google Health

- Package: google-health-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Google OAuth 2.0
- Summary: A local-first Google Health MCP connector that helps agents test the new Google Health API v4 without pasting Google OAuth tokens or health data into chat.
- Best for: Best for builders migrating Fitbit integrations, Pixel Watch users, and early Google Health API testers who want an agent-ready local OAuth workflow.
- Signals: Google Health API v4, Fitbit Migration, Steps, Sleep, Heart, Rollups
- Install: npx -y google-health-mcp-unofficial setup
- Doctor: npx -y google-health-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/google-health | GitHub: https://github.com/davidmosiah/google-health-mcp | npm: https://www.npmjs.com/package/google-health-mcp-unofficial | Provider docs: https://developers.google.com/health

Data availability:
- Identity, profile and settings: available. Uses users.getIdentity, users.getProfile and users.getSettings through Google Health API v4.
- Steps, activity, calories and distance: available. Available through supported data types and granted activity scopes.
- Sleep, heart rate, HRV, SpO2 and weight: conditional. Depends on supported devices, account data, data type support and granted scopes.
- Reconciled streams: available. Uses the API reconcile endpoint to query a single stream across sources where supported.
- Daily and physical-time rollups: available. Uses dailyRollUp and rollUp for supported data types.
- Writes, deletes, webhooks and TCX export: not_available. The public connector is intentionally read-only for the beta launch.
- Continuous live device telemetry: not_available. Not exposed as a live sensor stream by this MCP.

Agent tools:
- Profile and onboarding: `google_health_profile_get`, `google_health_profile_update`, `google_health_onboarding`
- Start here: `google_health_connection_status`, `google_health_data_inventory`, `google_health_daily_summary`, `google_health_weekly_summary`
- Auth and diagnostics: `google_health_capabilities`, `google_health_agent_manifest`, `google_health_privacy_audit`, `google_health_cache_status`, `google_health_get_auth_url`, `google_health_exchange_code`, `google_health_revoke_access`
- Identity and settings: `google_health_get_identity`, `google_health_get_profile`, `google_health_get_settings`
- Data points: `google_health_list_data_points`, `google_health_reconcile_data_points`
- Rollups and context: `google_health_daily_rollup`, `google_health_rollup`, `google_health_wellness_context`

Privacy:
- Google OAuth tokens stay local under ~/.google-health-mcp.
- The MCP client never receives access tokens, refresh tokens or client secrets.
- The default privacy mode is structured, with raw upstream JSON only by explicit request.
- The connector is read-only while the Google Health API v4 launch window is still moving.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Google documents possible breaking changes until the end of May 2026.
- Requires a Google Cloud OAuth client and user consent.
- Data availability depends on Google Health API support, account data and granted scopes.
- Not affiliated with Google, Fitbit or Alphabet.

### Withings

- Package: withings-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0 signed
- Summary: A Withings connector that gives agents body composition, sleep, activity, workouts and device-backed heart context.
- Best for: Best for body-composition, sleep and cardiometabolic wellness agents that need smart-scale and Withings device records.
- Signals: Body Measures, Sleep, Activity, Heart Records
- Install: npx -y withings-mcp-unofficial setup
- Doctor: npx -y withings-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/withings | GitHub: https://github.com/davidmosiah/withingsmcp | npm: https://www.npmjs.com/package/withings-mcp-unofficial | Docs: https://withingsmcp.vercel.app/ | Provider docs: https://developer.withings.com/api-reference/

Data availability:
- Body measures: available. Weight, fat percentage, muscle, bone and water with user.metrics scope.
- Daily activity: available. Steps, calories, distance and intensity.
- Workouts and sport metadata: available. Requires user.activity scope.
- Sleep summaries and detail records: available. Depends on device support.
- Heart records: conditional. ECG, blood pressure and other records vary by device/plan.
- Continuous sensor telemetry: not_available. Not exposed by Withings Public API.

Agent tools:
- Profile and onboarding: `withings_profile_get`, `withings_profile_update`, `withings_onboarding`
- Start here: `withings_connection_status`, `withings_daily_summary`, `withings_weekly_summary`
- Auth and diagnostics: `withings_capabilities`, `withings_agent_manifest`, `withings_privacy_audit`, `withings_cache_status`, `withings_get_auth_url`, `withings_exchange_code`, `withings_revoke_access`
- Body and metrics: `withings_list_body_measures`, `withings_list_heart`
- Activity: `withings_list_activity`, `withings_list_workouts`
- Sleep: `withings_list_sleep_summary`, `withings_list_sleep`

Privacy:
- OAuth tokens stay local.
- Signed API flow is handled inside the connector.
- Heart and body records are only available when the user's devices and scopes expose them.
- Agents can use summaries before requesting detailed records.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Data varies by Withings device and plan.
- OAuth authorization codes are short-lived.
- No raw continuous telemetry.

### Apple Health

- Package: apple-health-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Local export.zip
- Summary: A local parser for Apple Health exports that lets MCP clients summarize activity, sleep, heart, HRV, workouts and body records.
- Best for: Best for users who want Apple Health context in an agent without relying on a live API, cloud sync or background mobile service.
- Signals: Activity, Sleep, Heart Rate, HRV, Workouts
- Install: npx -y apple-health-mcp-unofficial setup --export-path /path/to/export.zip
- Doctor: npx -y apple-health-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/apple-health | GitHub: https://github.com/davidmosiah/apple-health-mcp | npm: https://www.npmjs.com/package/apple-health-mcp-unofficial | Provider docs: https://support.apple.com/guide/iphone/share-health-and-fitness-data-iph27f6325b2/ios

Data availability:
- Activity: available. Steps, distance, energy and exercise records.
- Heart rate: available. Recorded samples and resting heart rate.
- Sleep analysis and stages: available. When iPhone or Apple Watch logs sleep.
- Workouts and sport metadata: available. All HKWorkout entries in the export.
- Body measurements: available. Weight, BMI and body fat when logged.
- HRV and breathing rate: available. Available when Apple Watch supports and records them.
- ECG waveform: conditional. Metadata is available; raw waveform requires PDF export.
- Live HealthKit or iCloud Health sync: not_available. Not exposed by export files.

Agent tools:
- Profile and onboarding: `apple_health_profile_get`, `apple_health_profile_update`, `apple_health_onboarding`
- Start here: `apple_health_connection_status`, `apple_health_daily_summary`, `apple_health_weekly_summary`
- Diagnostics: `apple_health_capabilities`, `apple_health_agent_manifest`, `apple_health_privacy_audit`
- Records: `apple_health_list_records`, `apple_health_list_workouts`

Privacy:
- No OAuth or cloud token is required.
- All parsing happens from a local export path.
- The agent sees MCP responses, not the user's entire export by default.
- Good fit for private, server-side Hermes deployments when the export file is controlled.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Not live HealthKit.
- Requires a manual export refresh.
- Very large exports can take time to parse.

### Samsung Health

- Package: samsung-health-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Local CSV/ZIP export
- Summary: A local parser for Samsung Health personal-data downloads that lets MCP clients summarize Galaxy Watch activity, sleep, heart, HRV, body and workout records.
- Best for: Best for Galaxy Watch users who want Samsung Health context in an agent without handing over a Samsung account or waiting for an Android bridge.
- Signals: Galaxy Watch, Sleep, Heart Rate, HRV, Workouts
- Install: npx -y samsung-health-mcp-unofficial setup --export-path /path/to/SamsungHealth
- Doctor: npx -y samsung-health-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/samsung-health | GitHub: https://github.com/davidmosiah/samsung-health-mcp | npm: https://www.npmjs.com/package/samsung-health-mcp-unofficial | Provider docs: https://www.samsung.com/us/support/answer/ANS10001379/

Data availability:
- Steps and activity: available. Parsed from Samsung Health CSV filenames and headers.
- Heart rate, resting heart rate and HRV: available. When exported by the device/account.
- Sleep and sleep stages: available. When Galaxy Watch sleep records are present.
- Workouts and exercise records: available. Duration, distance, calories and activity type when exported.
- Body metrics: available. Weight and body-fat records when logged or synced.
- Live Health Connect or Samsung Health Data SDK reads: not_available. Planned as a separate Android bridge, not this Node MCP.

Agent tools:
- Profile and onboarding: `samsung_health_profile_get`, `samsung_health_profile_update`, `samsung_health_onboarding`
- Start here: `samsung_health_connection_status`, `samsung_health_data_inventory`, `samsung_health_daily_summary`, `samsung_health_weekly_summary`
- Diagnostics: `samsung_health_capabilities`, `samsung_health_agent_manifest`, `samsung_health_privacy_audit`
- Records: `samsung_health_list_records`, `samsung_health_list_workouts`

Privacy:
- No Samsung account credential is required.
- All parsing happens from a local folder, CSV file or zip.
- The agent sees bounded MCP responses, not the user's entire export by default.
- Summary mode is the default because Samsung Health exports are sensitive.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Not live Health Connect.
- Requires a manual export refresh.
- CSV columns can vary by app version, locale and device.

### Polar

- Package: polar-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: OAuth 2.0
- Summary: A Polar MCP connector for agents that need training, recovery, Nightly Recharge, PPI/HRV and physiology context from AccessLink.
- Best for: Best for training agents that need Polar recovery, sleep, continuous samples, route-aware sessions and test results.
- Signals: Nightly Recharge, Training Sessions, PPI Samples, Sleep, Continuous HR
- Install: npx -y polar-mcp-unofficial setup
- Doctor: npx -y polar-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/polar | GitHub: https://github.com/davidmosiah/polarmcp | npm: https://www.npmjs.com/package/polar-mcp-unofficial | Docs: https://polarmcp.vercel.app/ | Provider docs: https://www.polar.com/polar-api-v4/

Data availability:
- Daily activity and calendar: available. Requires activity:read and calendar:read.
- Sleep and sleep/wake vectors: available. Requires sleep:read.
- Nightly Recharge: available. Requires nightly_recharge:read and supported devices.
- Training sessions and targets: available. Requires training session/target scopes.
- Continuous samples and PPI samples: available. Requires continuous_samples:read and ppi_data:read.
- Routes and GPS geometry: opt_in. GPS coordinates are redacted unless raw mode is requested.
- Live device telemetry: not_available. Not exposed by Polar AccessLink.

Agent tools:
- Profile and onboarding: `polar_profile_get`, `polar_profile_update`, `polar_onboarding`
- Start here: `polar_connection_status`, `polar_daily_summary`, `polar_weekly_summary`
- Auth and diagnostics: `polar_capabilities`, `polar_agent_manifest`, `polar_privacy_audit`, `polar_cache_status`, `polar_get_auth_url`, `polar_exchange_code`, `polar_revoke_access`
- Account: `polar_get_account_data`, `polar_list_user_devices`, `polar_list_subscriptions`
- Activity and sleep: `polar_list_activity`, `polar_list_calendar`, `polar_list_sleeps`, `polar_list_sleep_wake_vectors`, `polar_list_nightly_recharge`
- Heart, physiology and training: `polar_list_continuous_samples`, `polar_list_ppi_samples`, `polar_list_temperature_measurements`, `polar_list_skin_contacts`, `polar_list_training_sessions`, `polar_list_training_targets`, `polar_list_tests`
- Sports and routes: `polar_list_sports`, `polar_list_sport_profile_catalog`, `polar_list_sport_profiles`, `polar_get_route`

Privacy:
- OAuth tokens stay local.
- GPS coordinates are redacted unless raw mode is explicitly requested.
- The connector exposes only scopes granted by the user.
- Start with daily/weekly summary before requesting detailed samples.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Some data requires supported Polar devices.
- GPS is opt-in.
- No live device telemetry.

### Eight Sleep

- Package: eight-sleep-mcp-unofficial
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Mobile-app password grant
- Summary: A local-first MCP connector that lets agents read your Eight Sleep nightly score, smart-temperature program, alarms and adjustable-base state — and optionally tune the pod, gated behind an explicit mutation flag.
- Best for: Best for sleep-focused agents that want a single source of truth across pod temperature, alarms and nightly score without copy-pasting from the Eight Sleep app.
- Signals: Sleep Trends, Temperature Program, Alarms, Adjustable Base
- Install: npx -y eight-sleep-mcp-unofficial setup
- Doctor: npx -y eight-sleep-mcp-unofficial doctor
- Links: Page: https://wellness.delx.ai/connectors/eight-sleep | GitHub: https://github.com/davidmosiah/eight-sleep-mcp | npm: https://www.npmjs.com/package/eight-sleep-mcp-unofficial | Provider docs: https://www.eightsleep.com/

Data availability:
- User profile and assigned bed side: available. Email and device serial redacted by default.
- Smart-temperature schedule (bedtime / initial sleep / final sleep): available. Includes currentLevel, currentDeviceLevel and isOn.
- Sleep trends (nightly score, stages, tnt, presence): available. Date-range query against client-api.8slp.net.
- Alarms and recommended next alarm: available. Includes thermal + vibration toggles.
- Adjustable base angle and preset: available. Returns leg/torso angles when the pod has the adjustable base.
- Write tools (temperature, side, away mode, alarm snooze/dismiss): opt_in. Off by default; require EIGHT_SLEEP_ALLOW_MUTATIONS=true.
- Continuous biometric streams or BLE: not_available. Not exposed by the mobile-app API surface.

Agent tools:
- Start here: `eight_sleep_connection_status`, `eight_sleep_get_me`, `eight_sleep_nightly_summary`, `eight_sleep_wellness_context`
- Reads: `eight_sleep_get_user`, `eight_sleep_get_current_device`, `eight_sleep_get_temperature`, `eight_sleep_get_alarms`, `eight_sleep_get_trends`, `eight_sleep_get_base`
- Workflows: `eight_sleep_nightly_summary`, `eight_sleep_wellness_context`
- Mutations (gated): `eight_sleep_set_temperature`, `eight_sleep_set_side`, `eight_sleep_set_away_mode`, `eight_sleep_snooze_alarm`, `eight_sleep_dismiss_alarm`
- Meta and diagnostics: `eight_sleep_capabilities`, `eight_sleep_agent_manifest`, `eight_sleep_data_inventory`, `eight_sleep_privacy_audit`, `eight_sleep_cache_status`, `eight_sleep_logout`

Privacy:
- Email/password stored locally in ~/.eight-sleep-mcp/config.json (chmod 600).
- Bearer token stored locally in ~/.eight-sleep-mcp/tokens.json (chmod 600).
- Email, device serial, addresses and payment last-four redacted from responses by default.
- Mutations require EIGHT_SLEEP_ALLOW_MUTATIONS=true — agents cannot accidentally change pod state.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 15 Delx Wellness connectors and never uploaded.

Known limits:
- Unofficial — Eight Sleep can change mobile-app endpoints without notice.
- Password auth is needed because no public OAuth grant exists.
- No continuous biometric streams or BLE access.

### Exercise Catalog

- Package: exercise-catalog-mcp-private
- Visibility: private_lab
- Readiness tier: experimental
- Auth model: Private provider token
- Summary: A private MCP connector for exploring how agents can combine wearable readiness with exercise selection and Telegram-ready recommendations.
- Best for: Best for internal Delx/Hermes testing where wearable signals can be combined with a controlled exercise database.
- Signals: Exercises, Muscle Groups, Equipment, Video Guidance, Workout Builder
- Install: Private lab only
- Doctor: Private lab only
- Links: Page: https://wellness.delx.ai/connectors/exercise-catalog

Data availability:
- Exercise search and catalog metadata: available. Private provider-dependent data.
- Muscle groups and equipment filters: available. Used for agent workout construction.
- Video guidance links: conditional. Only when the provider returns public video URLs.
- Medical or injury diagnosis: not_available. The connector is not a clinical decision system.

Agent tools:
- Start here: `exercise_catalog_connection_status`, `exercise_catalog_capabilities`, `exercise_catalog_build_workout`
- Catalog: `exercise_catalog_search`, `exercise_catalog_get_exercise`, `exercise_catalog_list_filters`
- Agent readiness: `exercise_catalog_agent_manifest`, `exercise_catalog_privacy_audit`

Privacy:
- Private lab by default.
- No provider endpoint is published on the marketing site.
- Rate limits should protect the upstream service.
- Agent output should remain non-medical and user-goal oriented.

Known limits:
- Private lab, not public npm.
- Provider terms and rate limits must be respected.
- Not medical advice.

### Wellness Air

- Package: wellness-air
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Public + optional API token
- Summary: Closes the room-quality gap in the wellness loop. Pair AQI / PM2.5 / CO₂ with WHOOP recovery, Oura sleep, or Garmin training to explain unexplained dips. v0.1 ships full AirGradient public + owned support.
- Best for: Best for users who already track recovery + nutrition and want to factor environmental quality (bedroom CO₂, outdoor AQI) into agent reasoning.
- Signals: AQI, PM2.5, PM10, CO2, TVOC, NOx, Temperature, Humidity
- Install: npx -y wellness-air doctor
- Doctor: npx -y wellness-air doctor
- Links: Page: https://wellness.delx.ai/connectors/air | GitHub: https://github.com/davidmosiah/wellness-air | npm: https://www.npmjs.com/package/wellness-air | Provider docs: https://api.airgradient.com/public/docs/api/v1/

Data availability:
- AQI (computed from PM2.5): available. US EPA breakpoint formula.
- PM2.5, PM10: available. All providers.
- CO2: available. AirGradient, AirThings, Awair.
- TVOC, NOx: available. AirGradient.
- Temperature, humidity: available. All major providers.
- Hourly trend chart: opt_in. Lands in v0.2 with rolling aggregations.
- Webhook threshold alerts: not_available. Roadmap v0.4.

Agent tools:
- Profile and onboarding: `air_profile_get`, `air_profile_update`, `air_onboarding`
- Start here: `air_quickstart`, `air_demo`, `air_aqi_check`
- Diagnostics: `air_connection_status`, `air_capabilities`, `air_privacy_audit`, `air_data_inventory`
- Reads: `air_current_reading`, `air_daily_summary`, `air_compare_locations`, `air_search_public_sensors`

Privacy:
- Public AirGradient sensor reads require no auth.
- Owned-sensor tokens stay on your machine.
- No telemetry — wellness-air never phones home.
- Read-only by design.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- AirThings/PurpleAir/IQAir/Awair adapters land in v0.2.
- Daily aggregation is a single-snapshot in v0.1.

### Wellness Cycle Coach

- Package: wellness-cycle-coach
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Stateless (no auth)
- Summary: Built so AI agents finally serve the 50% of users they've ignored — and never persists cycle data. The agent passes period start dates in tool args, the coach returns phase + recommendations, the data evaporates.
- Best for: Anyone with a menstrual cycle who wants their AI agent to give phase-aware guidance without trusting another period-tracker SaaS with their data.
- Signals: Phase, Cycle day, Cycle length, Next period, Phase nutrition, Phase training, Phase hydration
- Install: npx -y wellness-cycle-coach doctor
- Doctor: npx -y wellness-cycle-coach doctor
- Links: Page: https://wellness.delx.ai/connectors/cycle-coach | GitHub: https://github.com/davidmosiah/wellness-cycle-coach | npm: https://www.npmjs.com/package/wellness-cycle-coach

Data availability:
- Current phase detection: available. menstrual / follicular / ovulatory / luteal.
- Cycle length estimation: available. Rolling average from passed-in history.
- Next-period prediction: available. With confidence (low/medium/high) based on history depth.
- Phase nutrition guidance: available. Evidence-informed defaults per phase.
- Phase training intensity: available. Cycle-aware training-load recommendations.
- Phase hydration target: available. Per-phase ml/day target.
- Symptom-aware adjustments: opt_in. Lands in v0.3.

Agent tools:
- Profile and onboarding: `cycle_profile_get`, `cycle_profile_update`, `cycle_onboarding`
- Start here: `cycle_quickstart`, `cycle_demo`, `cycle_full_report`
- Diagnostics: `cycle_connection_status`, `cycle_capabilities`, `cycle_privacy_audit`, `cycle_data_inventory`
- Reasoning: `cycle_estimate_phase`, `cycle_predict_next_period`, `cycle_phase_guidance`, `cycle_recommend_nutrition`, `cycle_recommend_training`

Privacy:
- Stateless — never persists cycle data.
- Treat cycle data as medical-record sensitive.
- No outbound calls; pure-function computation.
- Defer to clinician for fertility, pregnancy, or symptom-management decisions.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- Stateless by design — agent must pass history each call.
- Recommendations are evidence-informed defaults, not medical guidance.
- PCOS/perimenopause edge cases roadmapped for v0.3.

### Wellness CGM

- Package: wellness-cgm-mcp
- Visibility: public
- Readiness tier: agent_ready
- Auth model: Dexcom OAuth + mock mode
- Summary: Pair cgm_meal_response with wellness-nourish meal logs to get 'this meal hit a 72 mg/dL spike at 45min — try protein-first next time'. Levels charges $199/mo for this; here it's an MCP. Sandbox mode lets any agent demo without setup.
- Best for: Biohackers, T2D users, athletes optimizing carb timing, anyone who wants their AI agent to see meal→glucose response in real time.
- Signals: Glucose, TIR (70-180), TIR (70-140), GMI (estimated A1C), CV, Meal response band
- Install: npx -y wellness-cgm-mcp doctor
- Doctor: npx -y wellness-cgm-mcp doctor
- Links: Page: https://wellness.delx.ai/connectors/cgm | GitHub: https://github.com/davidmosiah/wellness-cgm-mcp | npm: https://www.npmjs.com/package/wellness-cgm-mcp | Provider docs: https://developer.dexcom.com/docs/dexcomv2/overview

Data availability:
- Live EGV (5-minute glucose): available. Dexcom v3 API; mock fallback otherwise.
- Time-in-range (ADA 70-180): available. Computed from EGVs.
- Time-in-range (Levels 70-140): available. Metabolic-health profile for non-DM users.
- GMI (estimated A1C): available. Bergenstal 2018 formula: 3.31 + 0.02392 × mean(mg/dL).
- CV (variability): available. stdev / mean × 100.
- Meal response (peak + band): available. Pair with wellness-nourish meal logs.
- Refresh-token rotation: opt_in. Lands in v0.2.
- FreeStyle Libre (LibreLink Up): not_available. Roadmapped for v0.2.

Agent tools:
- Profile and onboarding: `cgm_profile_get`, `cgm_profile_update`, `cgm_onboarding`
- Start here: `cgm_quickstart`, `cgm_demo`, `cgm_glucose_now`
- Diagnostics: `cgm_connection_status`, `cgm_capabilities`, `cgm_privacy_audit`, `cgm_data_inventory`
- Reasoning: `cgm_glucose_window`, `cgm_daily_summary`, `cgm_meal_response`, `cgm_authorize_url`

Privacy:
- OAuth tokens stored in env, never returned in tool output.
- Mock mode clearly tagged mock=true on every payload.
- CGM data is medical-record sensitive — agent rules enforce that.
- Defer insulin/medication decisions to a clinician.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.

Known limits:
- v0.1 requires Dexcom only (Libre adapter in v0.2).
- Refresh-token rotation manual in v0.1.
- Not a replacement for clinical CGM workflows.

## Wellness modules

### Delx Wellness for Hermes

- Package: delx-wellness-hermes
- Role: Wellness agent profile
- Summary: Delx Wellness for Hermes turns the agent you already use into a recovery-aware wellness companion. It installs a dedicated delx-wellness profile, onboarding, wellness skills, connector presets for wearable and nutrition MCPs, and doctor checks without forking Hermes or creating a hosted token vault.
- Install: npx -y delx-wellness-hermes setup
- Doctor: npx -y delx-wellness-hermes doctor --profile delx-wellness
- Profile/runtime setup: npx -y delx-wellness-hermes setup
- Links: Page: https://wellness.delx.ai/hermes | GitHub: https://github.com/davidmosiah/delx-wellness-hermes | npm: https://www.npmjs.com/package/delx-wellness-hermes

Workflows:
- Run one setup command to create the delx-wellness Hermes profile.
- Start immediately with Nourish because it can smoke-test locally without OAuth.
- Configure a model/provider for chat readiness only after the profile and MCP layer are installed.
- Complete onboarding for goals, devices, training, nutrition and safety preferences before coaching.
- Use Telegram with Hermes as the first human UX for daily recovery and training decisions.
- Add WHOOP, Garmin, Oura, Google Health and other provider credentials through local connector setup flows; use local export paths for Apple Health and Samsung Health.

Agent tools:
- `delx-wellness-hermes setup`
- `delx-wellness-hermes install`
- `delx-wellness-hermes onboarding`
- `delx-wellness-hermes doctor`
- `Hermes profile model setup`
- `Hermes external skills`
- `Hermes mcp_servers presets`

Privacy:
- The package contains templates, skills and connector presets only.
- It does not contain user tokens, OAuth credentials, Telegram secrets or hosted hub API keys.
- Hosted hub mode requires an explicit --hub-url and never defaults to the public marketing site.
- The onboarding worksheet is local to the Hermes profile.

Known limits:
- Requires Hermes to be installed separately.
- Chat readiness requires a model/provider configured in Hermes.
- Connector credentials are still configured by each connector package.
- Exercise Catalog is disabled by default because private catalog access may depend on non-public data.
- The agent provides wellness context, not medical advice.

### Delx Wellness for OpenClaw

- Package: delx-wellness-openclaw
- Role: OpenClaw profile pack
- Summary: Delx Wellness for OpenClaw turns OpenClaw into a recovery-aware wellness agent. It installs a dedicated delx-wellness profile under ~/.openclaw-delx-wellness, writes OpenClaw-native mcp.servers presets for wearable and nutrition MCPs, adds SOUL.md, AGENTS.md, onboarding and wellness skills, and validates setup without creating a hosted token vault.
- Install: npx -y delx-wellness-openclaw setup
- Doctor: npx -y delx-wellness-openclaw doctor --profile delx-wellness
- Profile/runtime setup: npx -y delx-wellness-openclaw setup
- Links: Page: https://wellness.delx.ai/openclaw | GitHub: https://github.com/davidmosiah/delx-wellness-openclaw | npm: https://www.npmjs.com/package/delx-wellness-openclaw

Workflows:
- Run one setup command to create the delx-wellness OpenClaw profile.
- Validate OpenClaw config with openclaw --profile delx-wellness config validate.
- Check MCP visibility with openclaw --profile delx-wellness mcp list and mcp show nourish --json.
- Complete onboarding for goals, devices, training, nutrition and safety preferences before coaching.
- Start local agent turns with openclaw --profile delx-wellness agent --local --message.
- Add WHOOP, Garmin, Oura, Google Health and other provider credentials through local connector setup flows; use local export paths for Apple Health and Samsung Health.

Agent tools:
- `delx-wellness-openclaw setup`
- `delx-wellness-openclaw install`
- `delx-wellness-openclaw onboarding`
- `delx-wellness-openclaw doctor`
- `OpenClaw profile model setup`
- `OpenClaw skills.load.extraDirs`
- `OpenClaw mcp.servers presets`

Privacy:
- The package contains templates, skills and connector presets only.
- It does not contain user tokens, OAuth credentials, chat-channel secrets or hosted hub API keys.
- Hosted hub mode requires an explicit --hub-url and never defaults to the public marketing site.
- The onboarding worksheet is local to the OpenClaw profile workspace.

Known limits:
- Requires OpenClaw to be installed separately.
- Chat readiness requires a model/provider configured in OpenClaw.
- Connector credentials are still configured by each connector package.
- Exercise Catalog is disabled by default because private catalog access may depend on non-public data.
- The agent provides wellness context, not medical advice.

### Nourish MCP

- Package: wellness-nourish
- Role: Nutrition module
- Summary: Nourish is the Delx Wellness nutrition layer. It is a local-first MCP server for calories, macros, TACO/BR food search, barcode image decoding, meal photo previews, carbon context and nutrition trends, built to sit beside wearable connectors instead of pretending to be another device provider.
- Install: npx -y wellness-nourish setup --client claude
- Doctor: npx -y wellness-nourish doctor
- Profile/runtime setup: npx -y wellness-nourish setup --client hermes --profile personal --local-dir /root/.hermes/nourish/personal
- Links: Page: https://wellness.delx.ai/nutrition | GitHub: https://github.com/davidmosiah/wellness-nourish | npm: https://www.npmjs.com/package/wellness-nourish

Workflows:
- Preview a meal estimate before writing to the local log.
- Install Hermes personal mode for Telegram with setup --client hermes.
- Search TACO/BR local food data for Brazilian staples before falling back to USDA.
- Search USDA FoodData Central before estimating generic foods.
- Lookup packaged foods through Open Food Facts with ODbL attribution.
- Compare days, inspect trends, undo the last entry, or bulk-log a Telegram food recap.
- Summarize meal carbon footprint when food names have kg CO2e matches.
- Decode barcode photos from image paths, base64 images or data URIs before Open Food Facts lookup.
- Use agent vision observations for meal photo previews, then confirm portions before logging.
- Summarize calories, macros, hydration and goals in agent-friendly markdown.
- Export local intake data as JSONL or CSV.

Agent tools:
- `nourish_profile_get`
- `nourish_profile_update`
- `nourish_onboarding`
- `nourish_connection_status`
- `nourish_search_food`
- `nourish_lookup_barcode`
- `nourish_decode_barcode_image`
- `nourish_lookup_barcode_image`
- `nourish_estimate_meal`
- `nourish_estimate_meal_photo`
- `nourish_log_intake`
- `nourish_bulk_log_intake`
- `nourish_undo_last`
- `nourish_compare_days`
- `nourish_carbon_summary`
- `nourish_daily_summary`
- `nourish_weekly_summary`
- `nourish_log_water`
- `nourish_set_goals`
- `nourish_export_data`

Privacy:
- Food logs, hydration and goals stay under ~/.wellness-nourish unless NOURISH_LOCAL_DIR is set.
- Profile and goals persist locally to ~/.delx-wellness/profile.json (mode 0600); shared with all 14 Delx Wellness connectors and never uploaded.
- No hosted account is required.
- Agents must preserve source attribution, confidence and safety warnings.
- Mutating MCP tools require explicit user intent.

Known limits:
- Nutrition estimates are approximate.
- Meal photo estimates depend on the agent vision layer and always require user confirmation before logging.
- No hosted sync or clinical guidance.
- Open Food Facts data carries attribution and share-alike obligations.
- TACO values ship as an attributed curated subset while redistribution-license confirmation remains a 1.0 gate.

## Recommended agent behavior

1. Start from connection_status, agent_manifest, privacy_audit and summary tools before fetching detailed records.
2. Prefer daily and weekly summaries for wellness recommendations.
3. Ask permission before raw payloads, location data or provider-specific sensitive details.
4. Treat output as wellness context, not diagnosis, treatment or emergency monitoring.
5. Mention provider limitations when data is conditional, opt-in or unavailable.