🛡

Strava Integration Options for PKA

Technical Assessment & Recommendation

Prepared by: Neo, Chief Code Architect

Date: 2026-03-28

Status: For Will's review

The Elephant in the Room: Strava's AI Prohibition

Before evaluating options, you need to know this. In November 2024, Strava updated their API agreement with an explicit clause:

"You may not use the Strava API Materials (including Strava Data), directly or indirectly, for any model training related to artificial intelligence, machine learning or similar applications."

They also added:

"You may only display or disclose to an end user the specific Strava Data related to that user."

There is no personal-use exception in the agreement. Feeding your own Strava API data into Claude for analysis technically falls under "use in AI applications."

Practical reality: Strava is not going to detect or enforce against a single user querying their own data through a personal API app and discussing it with Claude. Multiple MCP servers exist openly on GitHub doing exactly this. But the legal footing is not clean. Your bulk-exported data (GDPR download) is yours outright — no API agreement applies to it.

Option 1: Strava MCP Server (Direct API Integration)

Multiple open-source MCP servers exist that connect Claude directly to your Strava account via OAuth 2.0. The strongest:

Server	Language	Tools	Stars	Maturity
r-huijts/strava-mcp	TypeScript/Node.js	25 tools	Active	Most complete
MariyaFilippova/mcp-strava	Kotlin/JVM	18 tools	18	Route suggestions, Google Maps links

What It Unlocks

All your activities — distance, time, speed, power, HR zones, cadence, laps, GPS streams
Athlete stats — YTD totals, all-time PRs, activity type breakdowns
Segments — starred segments, nearby segment discovery, leaderboards, your efforts
Routes — saved routes, GPX/TCX export
Clubs — membership info, club activities
Real-time conversational analysis — "How was my ride today?" "Compare my March vs February cycling."

What It Costs

Money: Free. Strava API is free for personal apps.
Effort: 15 minutes to set up.
Rate limits: 200 requests per 15 minutes, 2,000 per day.
Legal risk: Low-but-nonzero. Technically violates the AI clause. Practically unenforced for personal use.

Recommendation: Do this first. Takes 15 minutes. Immediate payoff.

Option 2: Strava Bulk Export + Local Analysis

Strava lets you download your complete data archive via Settings > My Account > Download or Delete Your Account > Request Your Archive. You receive a ZIP file within 2-4 hours containing:

activities.csv — every activity with metadata
Individual .gpx / .fit / .fit.gz files for every activity with full GPS tracks
profile.csv, goals.csv, and other account data

This data is yours under GDPR. No API agreement applies. You can do whatever you want with it.

Recommendation: Do this regardless of other options. It is your data backup and your legally clean dataset for deep analysis.

Option 3: Chrome MCP / Browser Scraping

What It Unlocks

Anything visible in the browser — dashboard, activity feed, segment leaderboards, heatmap, clubs
Data that the API restricts — other users' data visible in browser
Segment intelligence — athlete counts, effort counts, popular times
Heatmap data — browser-only, not available via API

Recommendation: Use selectively for data the API cannot provide (heatmaps, segment crowd intel). Not worth building a full pipeline — too fragile.

Option 4: Hybrid Architecture (Recommended)

Layer 1: Strava MCP Server (real-time)
  - Daily/conversational access to your recent activities
  - Quick stats, comparisons, training summaries
  - "How was my ride?" workflows

Layer 2: Bulk Export (periodic, legally clean)
  - Full historical archive in mainframe
  - activities.csv -> strava_activities table
  - GPS tracks stored locally for route analysis
  - Re-export quarterly or before major analysis

Layer 3: Chrome MCP (targeted, on-demand)
  - Segment crowd intelligence for trip planning
  - Heatmap analysis for route discovery in new zones
  - Leaderboard data for competitive segments

Cross-Referencing with Existing PKA Data

Strava activities + Activity Zones — auto-tag each activity to a zone by GPS coordinates
Strava activities + RWGPS routes — match rides to planned routes, identify unplanned explorations
Strava segments + Route planning — use segment popularity as a quality signal
Strava training load + Trip scheduling — inform trip intensity planning based on recent fitness data

Implementation Plan

Step	Action	Time	Depends On
1	Request Strava bulk export	5 min	Nothing
2	Create Strava API app	10 min	Nothing
3	Install `r-huijts/strava-mcp` server	15 min	Step 2
4	Download bulk export, extract	10 min	Step 1 (wait 2-4 hrs)
5	Add `strava_activities` table to mainframe	30 min	Step 4
6	Write `sync_strava_bulk()`	1 hr	Step 5
7	Write `match_activity_to_zone()`	30 min	Step 6
8	Build segment intelligence scraping	2-3 hrs	As needed

Decision Summary

Option	Verdict	Why
Strava MCP Server	DO IT	15-minute setup, immediate conversational access to your training data.
Bulk Export	DO IT	Your data, your rules. No legal gray area. Foundation for mainframe integration.
Chrome MCP Scraping	USE SELECTIVELY	Good for segment intel and heatmaps. Not worth a full pipeline.
Composio/third-party	SKIP	Adds a middleman for no benefit.

Bottom line: Install the MCP server for daily use. Request the bulk export for deep analysis and mainframe integration. Use Chrome MCP for segment crowd intelligence when planning activities in new zones.

Standing by for your call.

-- Neo

Sources:

Voice Reply