io.github.davidmosiah/strava-mcp

Strava MCP

Give your AI agent your Strava activities, streams, segments and routes — locally.

Local-first MCP server — tokens never leave your machine.

⚡ 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.

HTTP (v2 stateless)

Default is stdio. Optional Streamable HTTP — no session id, JSON responses, loopback only:

npx -y strava-mcp-unofficial --http
# GET  http://127.0.0.1:3000/health
# POST http://127.0.0.1:3000/mcp   (sessionless)

Env: STRAVA_MCP_HOST, STRAVA_MCP_PORT, STRAVA_MCP_TRANSPORT=http.

Local-first MCP server that connects AI agents to your Strava activities, routes, streams and training context.

Unofficial project. Not affiliated with, endorsed by or supported by Strava, Inc. Strava is a trademark of its respective owner. Use this only with your own Strava account and in line with Strava's API agreement.

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

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

Strava holds the long memory of your training — every ride, run, swim, segment, route and stream. But it lives behind an OAuth API with strict rate limits (200 req/15min, 2k/day per app) and GPS data that's privacy-sensitive by default.

This package does the OAuth dance locally, throttles under Strava's per-app limits, redacts GPS lat/lng unless you explicitly opt in, and exposes Strava through the Model Context Protocol. Any MCP-compatible agent gets your training context with one config snippet. Tokens never leave your machine.

Quickstart

From zero to your first agent call in about a minute. You only need a Strava app (create one here) with redirect URI http://127.0.0.1:3000/callback.

1. Paste your app's client id + secret (interactive, stored at ~/.strava-mcp/config.json with 0600):

npx -y strava-mcp-unofficial setup

2. Authorize Strava. auth opens your browser; --no-open prints the URL so you can paste it yourself (handy on a headless box). Tokens are saved locally — the command never prints them:

$ npx -y strava-mcp-unofficial auth --no-open
Strava MCP · Authorization

Open this URL manually:
  https://www.strava.com/oauth/authorize?client_id=12345&redirect_uri=http%3A%2F%2F127.0.0.1%3A3000%2Fcallback&response_type=code&approval_prompt=auto&scope=read%2Cactivity%3Aread_all%2Cprofile%3Aread_all&state=aa38f29b

Steps
  1. Approve access in the browser tab that opens.
  2. Strava will redirect to the local callback.
  3. Tokens are saved locally; this command never prints them.

Waiting for callback...

3. Verify you're ready — doctor confirms scopes and setup without calling Strava:

$ npx -y strava-mcp-unofficial doctor
Strava MCP · Doctor
Status: READY ✓

Checks
  ✓  Node.js >=20
  ✓  Env vars
  ✓  Local config
  ✓  Automatic auth redirect
  ✓  Token file
  ✓  Token permissions
  ✓  Refresh token
  ✓  OAuth scopes
  ·  Privacy mode
  ·  Cache

Next steps
  1. Ready. Add this MCP server to your agent and start with strava_daily_summary.

If OAuth scopes shows a ✗, re-run auth and approve activity:read_all profile:read_all read.

4. Make a first call — no live account required. Ask your agent to run strava_demo. It returns realistic, synthetic payloads (tagged is_demo: true) so you can wire prompts before connecting real data:

> Call strava_demo and summarize my week.

# Strava Demo

- **is_demo**: true
- **recent_sessions**: 3
- **load_classification**: moderate
- **primary_signal**: Recent Strava load is manageable; use intent and consistency as the main lever.
- **recommended_handoff**: exercise_catalog_recommend_session

Call it with response_format: "json" for the full shapes. The sample block mirrors strava_daily_summary, strava_training_context and strava_list_activities key-for-key — npm run test:demo-contract runs the real builders over fixtures/strava-activities.mjs and fails the build if the demo ever invents a field or omits one. Swap strava_demo for the real tool and the same shape arrives filled with your Strava data.

5. Wire it into your MCP client:

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

For Claude Desktop, run setup --client claude and the snippet is written for you. For Hermes, see Hermes / remote setup below.

Try it with your agent

Three things to ask first:

Use strava_connection_status to check setup, then run strava_daily_summary.
Tell me what my training context looks like in 5 lines.

Call strava_weekly_summary with response_format=json. Find my biggest
load/intensity bottleneck and give me a next-week endurance plan.

Use the strava_activity_stream_investigator prompt for activity_id=<id>.
Don't expose GPS unless I explicitly ask for it.

Data availability

This package uses the official Strava API v3. When this README says raw, it means the upstream Strava JSON for a supported endpoint — not continuous device telemetry.

Tools

Start with these:

• strava_connection_status — verify local setup, scopes and readiness before calling Strava
• strava_data_inventory — inventory supported data domains, scopes, privacy modes and recommended first calls without calling Strava APIs.
• strava_daily_summary — latest activity, weekly load and intensity context for today
• strava_weekly_summary — scorecard, comparison vs prior week, next-week training plan

Auth & diagnostics

• strava_capabilities, strava_agent_manifest, strava_privacy_audit, strava_cache_status
• strava_get_auth_url, strava_exchange_code, strava_revoke_access

Athlete & training

• strava_get_athlete, strava_get_zones, strava_get_athlete_stats

Activities & streams

• strava_list_activities, strava_get_activity, strava_get_activity_zones
• strava_activity_series, strava_get_activity_streams — GPS lat/lng requires include_gps=true or raw mode

Routes & context

• strava_list_routes, strava_get_route, strava_list_clubs, strava_get_gear

Prompts

• strava_daily_training_director — practical daily training brief
• strava_weekly_endurance_review — week comparison + next-week endurance plan
• strava_activity_stream_investigator — investigate one activity using streams (GPS-aware)

Each accepts timezone (IANA, default UTC).

Resources

• strava://capabilities, strava://agent-manifest
• strava://athlete
• strava://latest/activity
• strava://summary/daily, strava://summary/weekly

Privacy & security

• OAuth tokens are stored in ~/.strava-mcp/tokens.json with 0600 permissions and are never returned by tools.
• Write/upload scopes are not requested by default — read-only by design.
• GPS lat/lng and route geometry are recursively removed in summary and structured modes, and only included with explicit include_gps=true for stream calls or raw mode.
• Structured mode otherwise preserves complete upstream physiological records and future Strava fields.
• Activity after / before filters retain instant semantics when converted from timezone-aware ISO date-times to Strava epoch seconds; invalid ranges fail before HTTP.
• Route geometry is also redacted unless raw mode is explicitly requested.
• 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 ~/.strava-mcp/config.json (0600). Manual env override is supported:

STRAVA_CLIENT_ID=…
STRAVA_CLIENT_SECRET=…
STRAVA_REDIRECT_URI=http://127.0.0.1:3000/callback

# Optional
STRAVA_SCOPES="read activity:read_all profile:read_all"
STRAVA_PRIVACY_MODE=structured        # summary | structured | raw
STRAVA_CACHE=sqlite                   # optional read-through cache

Hermes / remote setup

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

Hermes commonly exposes Strava tools with a prefix:

• mcp_strava_strava_agent_manifest
• mcp_strava_strava_connection_status
• mcp_strava_strava_daily_summary
• mcp_strava_strava_weekly_summary
• mcp_strava_strava_get_activity_streams

After Hermes config changes, use /reload-mcp or hermes mcp test strava. 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 ~/.strava-mcp/tokens.json to the server with chmod 600. The token must include activity:read_all profile:read_all read for activity history and streams.

Requirements

• Node.js 20+
• A Strava app with redirect URI http://127.0.0.1:3000/callback

Why these scopes:

• read — public profile, routes and public Strava resources
• activity:read_all — your activities, including private activities visible to your app
• profile:read_all — fuller authenticated athlete profile fields

No write scope is requested by default.

Development

git clone https://github.com/davidmosiah/strava-mcp.git
cd strava-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/strava-mcp-unofficial
• Docs site: https://wellness.delx.ai/connectors/strava
• Legacy docs: https://stravamcp.vercel.app/
• GitHub Pages mirror: https://davidmosiah.github.io/strava-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
• Strava API docs: https://developers.strava.com/docs/reference/
• Strava auth docs: https://developers.strava.com/docs/authentication/

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.

Skill or MCP

Same package, two doors. MCP registers tools on stdio/HTTP. The skill can drive the same tools through the CLI when the client has no MCP:

npx -y strava-mcp-unofficial call strava_connection_status --json '{}'

Copy skill/SKILL.md into your agent skills dir.