Strava Segment & Club Intel

Public Strava segment and club intelligence the locked-down API no longer hands to third parties - normalized into one schema, with geometry-derived climb metrics and popular-segments-by-area rollups.

In November 2024 Strava tightened its API: new apps default to one authorized user, broader access goes through a review/approval form, AI/aggregation use is banned, and a developer subscription gates the standard tier. The people that locks out - fitness-app builders, coaches, route/segment-analytics tools, endurance-gear marketers - are exactly who needs this data. This actor delivers the public segment/club surface that doesn't require an application: logged-out, public-only, and PII-minimized.

Responsible use (read this first). This actor reads only logged-out, PUBLIC Strava data. It does not log in, use any athlete credential, or collect individual athletes' names, IDs, photos, leaderboard positions, or any private activity/movement trace. Club output is aggregate (member count only). Segment geometry is the segment's own public road/trail shape, not a person. Use it for segment/route and area analytics, not to profile or track individuals. Respect Strava's Terms and applicable privacy law in your jurisdiction.

---

What it does

Mode	Input	Output
segment_detail	segment_ids (IDs or URLs)	Normalized segment records - name, sport, location, and distance / average + max grade / elevation gain-loss / climb category derived from Strava's own public geometry stream - plus automatic area rollups.
club_profile	club_ids (IDs or URLs)	Aggregate club records - name, sport, location, club type, member count, verified flag. No member identities, ever.
area_explore	bounds or center_lat/lng + radius_km	Best-effort bounding-box discovery. Strava's explore endpoint is auth-gated when logged out, so this fail-softs with a documented note unless you supply a warm browser session (browser_cdp_url); place-level rollups still run over any segment_ids you provide.

Why the metrics are derived

Since the Nov-2024 lockdown, the logged-out segment page no longer ships the numeric stats (effort/athlete counts, leaderboards) - those load behind a session. What is public is the segment's geometry stream (/stream/segments/{id}: the distance, altitude, and lat/lng arrays of the segment's shape). This actor computes distance, grade, elevation gain/loss, and a Strava-style climb category from that public geometry, and labels every computed value metrics_source: "public_stream" so you never mistake a derived grade for an official one.

Worked example (live-verified): segment 229781 "Hawk Hill" -> 2,658 m, 5.86% avg grade, +156 m, Cat 4. Segment 8109834 "Old La Honda" -> 5,048 m, 7.7% avg grade, +389 m, Cat 2.

---

Output

Each record carries record_type: segment, club, area_rollup, plus a responsible_use_note header and any fetch_error / explore_gated notes. Dataset views: overview (AI-agent skim), segments, clubs, areas.

{
  "record_type": "segment",
  "segment_id": "229781",
  "name": "Hawk Hill",
  "activity_type": "ride",
  "display_location": "Marin Headlands (GGNRA), California",
  "distance_m": 2658.5, "distance_mi": 1.65,
  "average_grade_pct": 5.86, "max_grade_pct": 9.2,
  "elevation_diff_m": 155.9, "total_elevation_gain_m": 155.9,
  "climb_category": 4, "climb_category_label": "Cat 4",
  "start_latlng": [37.8331, -122.483451],
  "metrics_source": "public_stream",
  "source_url": "https://www.strava.com/segments/229781"
}

---

Pricing (pay-per-event)

Event	Price	When
segment_record	$0.006	One normalized segment or aggregate club record
metric_enrichment	$0.012	Per segment enriched with geometry-derived metrics
area_rollup	$0.015	Per popular-segments-by-area aggregation
scheduled_delta_run	$0.050	Per scheduled monitor run (new/changed deltas)

---

Anti-bot escalation ladder

Strava's public pages usually answer the cheapest tier from a datacenter IP. On a real block the client escalates automatically:

1. httpx over the DATACENTER proxy.
2. curl_cffi with Chrome TLS impersonation over the RESIDENTIAL proxy (defeats JA3/TLS-fingerprint WAFs).
3. Playwright (patchright stealth) over RESIDENTIAL / a warm BROWSER_CDP_URL session - for JS/Cloudflare challenges and to reach the auth-gated explore endpoint with the operator's own session (this actor never handles credentials).
4. Fail-soft - a documented note, run still finishes SUCCEEDED.

Set BROWSER_CDP_URL to point the browser tier at a warm anti-detect browser. CAPTCHA-less by default.

---

Monitor mode

Run on a schedule (Apify Schedules) and the actor computes a delta vs the prior run - new/changed segments and club member-count trends (an aggregate, privacy-safe "leaderboard trend" signal) - and can post a Slack digest (slack_webhook_url). Adds scheduled_delta_run only on scheduled runs.

---

What this actor will NOT do

• No login, no athlete OAuth, no cookies carried in.
• No individual athlete records, names, IDs, photos, or follower graphs.
• No segment leaderboards (gated logged-out anyway) and no KOM/QOM holder identities.
• No personal routes or activities - those are an individual's movement trace and are deliberately not built.

These are hard product boundaries, not toggles.

---

MCP / AI agents

A companion mcp-strava-segment-intel server exposes these as MCP tools (get_segments, get_club, area_rollup, explore_segments) for Claude, Cursor, OpenAI, and LangChain, with x402 (USDC on Base) and Skyfire agentic-payment support. The dataset also ships an MCP-compatible JSON link (see Output tab).

Who it's for

Cycling/running businesses, coaches and route/segment-analytics tools, endurance-gear marketers, and AI agents doing route/area research - anyone shut out by the Nov-2024 API lockdown who needs the public segment/club layer without an application.