io.github.davidmosiah/polarmcp

⚡ One-command install with Delx Wellness for Hermes: npx -y delx-wellness-hermes setup — preconfigures this connector and the other 8 in a dedicated Hermes profile.

Or wire it standalone into Claude Desktop / Cursor / ChatGPT Desktop — see the install section below.

Local-first MCP server that connects AI agents to your Polar training, sleep, Nightly Recharge and continuous-sample data.

Unofficial project. Not affiliated with, endorsed by or supported by Polar Electro Oy. Polar is a trademark of its respective owner. Use this only with your own Polar account and in line with the Polar AccessLink API terms.

Built by David Mosiah for people who use Claude, Cursor, Hermes, OpenClaw or other MCP-compatible agents to think about training load, recovery and endurance - without copy-pasting numbers from Polar Flow.

Part of Delx Wellness, a registry of local-first wellness MCP connectors.

If this connector helps your agent workflow, please star the repo. Stars make the project easier for other AI builders to discover and help Delx keep shipping local-first wellness infrastructure.

Why this exists

Polar has one of the deepest training-physiology stacks among consumer wearables - Nightly Recharge, continuous samples, PPI (pulse-to-pulse intervals), training targets, sport profiles, orthostatic and fitness tests. The Polar AccessLink Dynamic API v4 exposes this data, but with 16 fine-grained OAuth scopes and a structure that's harder to navigate than typical consumer APIs.

This package handles the OAuth dance locally, normalizes responses across the v4 endpoints, redacts GPS by default, and exposes Polar through the Model Context Protocol. Tokens never leave your machine.

Setup in 60 seconds

You'll need a Polar AccessLink client (create one here) with redirect URI http://127.0.0.1:3000/callback.

npx -y polar-mcp-unofficial setup    # interactive: paste client id + secret
npx -y polar-mcp-unofficial auth     # opens browser, captures the OAuth code
npx -y polar-mcp-unofficial doctor   # verifies you're ready

Recommended scopes (request the ones matching the data you want):

activity:read calendar:read continuous_samples:read devices:read
nightly_recharge:read ppi_data:read profile:read routes:read
skin_contact:read sleep:read sports:read temperature_measurement:read
tests:read training_sessions:read training_targets:read user_subscription:read

Then add this to your MCP client config:

{
  "mcpServers": {
    "polar": {
      "command": "npx",
      "args": ["-y", "polar-mcp-unofficial"]
    }
  }
}

For Claude Desktop, run setup --client claude and the snippet is written for you.

Try it with your agent

Three things to ask first:

Use polar_connection_status to check setup, then run polar_daily_summary.
Give me a 5-line training brief for today.

Call polar_weekly_summary with response_format=json. Identify my biggest
training-load/recovery bottleneck and give me a next-week plan.

Use the polar_training_load_investigation prompt, after=2026-04-01.
Walk me through my recent training sessions + Nightly Recharge.

Data availability

This package uses the official Polar AccessLink Dynamic API v4. When this README says raw, it means the upstream Polar JSON for a supported endpoint - not raw device sensor streams.

Tools

Start with these:

• polar_connection_status - verify local setup, scopes and readiness before calling Polar
• polar_data_inventory — inventory supported data domains, scopes, privacy modes and recommended first calls without calling Polar APIs.
• polar_daily_summary - sleep, activity, Nightly Recharge and training brief for today
• polar_weekly_summary - scorecard, comparison vs prior week, next-week plan

Auth & 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 & sleep

• polar_list_activity, polar_list_calendar
• polar_list_sleeps, polar_list_sleep_wake_vectors
• polar_list_nightly_recharge

Heart & physiology (date range)

• polar_list_continuous_samples, polar_list_ppi_samples
• polar_list_temperature_measurements, polar_list_skin_contacts

Training

• polar_list_training_sessions, polar_list_training_targets, polar_list_training_target_favorites
• polar_list_tests

Sports & routes

• polar_list_sports, polar_list_sport_profile_catalog, polar_list_sport_profiles
• polar_get_route - GPS coordinates redacted unless raw mode

Prompts

• polar_daily_checkin - practical daily training and recovery check-in
• polar_weekly_review - review trends across activity, sleep and recovery
• polar_training_load_investigation - investigate training sessions + recovery context

Resources

• polar://capabilities, polar://agent-manifest
• polar://summary/daily, polar://summary/weekly

Privacy & security

• OAuth tokens are stored in ~/.polar-mcp/tokens.json with 0600 permissions and are never returned by tools.
• The server never prints access or refresh tokens.
• POLAR_PRIVACY_MODE defaults to structured. Raw Polar JSON is opt-in via raw mode or per-call override.
• GPS route geometry is redacted in summary and structured modes - only raw mode exposes raw coordinates.
• The MCP client never sees access or refresh tokens.
• This is not medical advice. The server exposes user-authorized data for personal AI workflows, not diagnosis or training prescription.

Configuration

setup writes most of these into ~/.polar-mcp/config.json (0600). Manual env override is supported:

POLAR_CLIENT_ID=<client-id>
POLAR_CLIENT_SECRET=<client-secret>
POLAR_REDIRECT_URI=http://127.0.0.1:3000/callback

# Optional
POLAR_SCOPES="activity:read calendar:read continuous_samples:read ..."
POLAR_PRIVACY_MODE=structured        # summary | structured | raw
POLAR_CACHE=sqlite                   # optional read-through cache
POLAR_TOKEN_PATH=~/.polar-mcp/tokens.json
POLAR_CACHE_PATH=~/.polar-mcp/cache.sqlite

Hermes / remote setup

npx -y polar-mcp-unofficial setup --client hermes --no-auth
npx -y polar-mcp-unofficial auth                      # run locally if browser auth is needed
npx -y polar-mcp-unofficial doctor --client hermes
hermes mcp test polar

After Hermes config changes, use /reload-mcp or hermes mcp test polar. Don't restart the gateway for normal data access.

If browser OAuth has to happen on a different machine than Hermes, run auth locally and copy ~/.polar-mcp/tokens.json to the server with chmod 600.

Requirements

• Node.js 20+
• A Polar AccessLink client at https://admin.polaraccesslink.com with redirect URI http://127.0.0.1:3000/callback

Development

git clone https://github.com/davidmosiah/polar-mcp.git
cd polar-mcp
npm install
npm test
npm run build

Test with MCP Inspector:

npx @modelcontextprotocol/inspector node dist/index.js

Links

• npm: https://www.npmjs.com/package/polar-mcp-unofficial
• Docs site: https://wellness.delx.ai/connectors/polar
• Legacy docs: https://polarmcp.vercel.app/
• GitHub: https://github.com/davidmosiah/polar-mcp
• Delx Wellness registry: https://github.com/davidmosiah/delx-wellness
• Connector quality standard: https://github.com/davidmosiah/delx-wellness/blob/main/docs/connector-quality-standard.md
• Polar AccessLink Dynamic API v4 docs: https://www.polar.com/polar-api-v4/

See also

The full Delx Wellness connector library:

One-command setup for Hermes — preconfigures every connector above plus wellness skills + onboarding: delx-wellness-hermes.

📧 Contact & Support

• 📨 support@delx.ai — general questions, integration help, partnerships
• 🐛 Bug reports / feature requests — GitHub Issues
• 🐦 Updates — @delx369 on X
• 🌐 Site — wellness.delx.ai

License

MIT - see LICENSE.

Disclaimer

This software is provided as-is. It is not a medical device, does not provide medical advice, and should not be used for diagnosis, treatment or training prescription. Always consult qualified professionals for medical or training concerns.