scher.nl
Personal project · October 2026

Strava insights via MCP

This page shows how much scher rides, how power develops over the years and which unpaved paths nearby are still waiting. Claude fetched the rides through the Strava MCP and the roads from OpenStreetMap through overpass-api.de. Small Python scripts did the calculations, and the results sit as fixed data behind every chart on this page.

Total ridden
—
Rides
—
Ridden in this year
—
Estimated FTP this year
—
At a glance

From Strava and OpenStreetMap to this page

From Strava and OpenStreetMap to this page Claude reads rides and power data from Strava through the Strava MCP. Claude Code on the laptop fetches roads and terrain from OpenStreetMap through overpass-api.de and puts them on Google Drive, where Claude picks them up. Claude does the calculations and puts the data on this page. Stravarides · power Strava MCP Claudefetch · calculate datasets Dashboardthis page OpenStreetMaproads · terrain overpass-api.de LaptopClaude Code Google Drive
The source

What the Strava MCP provides

The Strava MCP is a connector that lets Claude read your own Strava data directly. It is read-only, cannot change or post anything, sees only your own account and requires a Strava subscription. This is what it returns:

Activitieslist_activities

Per activity the name, sport, start time, bike, distance, moving and elapsed time, elevation gain, average and maximum speed, cadence, calories, relative effort, kudos and PRs. On request also a simplified route line and tags such as commute or trainer. Used for distance per year, outdoor versus virtual and the map.

Streamsget_activity_streams

Measurements throughout the ride: time, GPS position, heart rate, power, cadence, distance, altitude, speed, grade, moving or stopped, and temperature. Used for the power curve and the estimated FTP.

Performance per rideget_activity_performance

Average and maximum heart rate, average power and cadence, calories, perceived exertion, PRs, segments, best efforts and laps.

Zones and profileget_athlete_zones · get_athlete_profile

Heart rate zones, power zones, FTP and running pace zones, and a profile with location, weight, units and training focus.

Gearget_gear

Bikes and shoes with their total distance, and for bikes also weight and frame type.

Otherget_club_info · get_training_plan · get_strength_workout_details

Clubs and their upcoming events, running training plans and the exercises of strength workouts.

What it doesn't include

The activity list has no average power, so for the power curve the power stream was fetched for each ride. Strava limits how many of those requests can be made in a short time. Roads and surfaces are not in Strava; they come from OpenStreetMap.

The method

How this page was made

  1. 1

    Fetching rides

    Strava MCP

    Claude requested all activities, including the simplified route lines, and saved them as files so they never need to be fetched again. Only rides count.

    list_activitiesroute lines
  2. 2

    Power per ride

    Strava MCP

    For every ride with a power meter the full power stream was fetched. Each year shows the highest average held within a single ride, from 5 seconds to an hour; FTP is 95% of the best 20 minutes.

    get_activity_streamsPython
  3. 3

    Roads from OpenStreetMap

    overpass-api.de

    Unpaved roads, cycling and MTB routes, closed-off areas and the background map come from OpenStreetMap, queried from the Overpass service at overpass-api.de. That service could not be reached from the cloud, so Claude Code on the laptop fetched the data and handed it over through Google Drive.

    Overpass QLClaude CodeGoogle Drive
  4. 4

    Combining

    Claude

    Python scripts laid the route lines on a 40 by 40 metre grid and sampled every unpaved road every 15 metres. That way each stretch of road is known to have been ridden or not.

    Python40 m grid
  5. 5

    Dashboard

    Claude dashboard

    The results first went into a Claude dashboard, where every number shows which data it comes from. This page is a fixed copy of it, with the data inside the page itself. The map is centred on the centre of Oisterwijk and only shows paths that have never been ridden.

    d3canvas
Distance

Distance ridden per year

All rides, outdoor and virtual together, in kilometres.

Outdoor or indoor

Outdoor versus virtual

Share of kilometres per year. Virtual is a ride of type VirtualRide.

OutdoorVirtual
Power

Power curve per year

Highest average power held per year, from 5 seconds to 1 hour, across all rides with power data. The dotted line in the same colour is that year's estimated FTP: 95% of the best 20 minutes. Click a year to show or hide it.

Unpaved

Unpaved paths not yet ridden

This map shows in orange the unpaved roads and paths within a 38 by 38 km square around Oisterwijk that scher has never ridden by bike. Drag to pan, scroll or use + and − to zoom.

The rides. The activities were fetched through the Strava MCP, with the simplified route line of every outdoor ride (Ride, GravelRide and related types; virtual rides do not count). Those lines were laid on a 40 by 40 metre grid, so each cell records which rides passed through it.

The roads. All map data comes from OpenStreetMap, through the Overpass service: unpaved roads and farm tracks, cycling and MTB routes, closed-off areas, and for the background roads, rail, water, forest and built-up areas. The map keeps only unpaved roads that can be ridden on a gravel bike (gravel, fine gravel, compacted, dirt and ground). Left out are sand, grass, mud and woodchips, soft tracks, walking paths and footpaths, paths where cycling is forbidden, roads only for agriculture, forestry, customers or with a permit, MTB routes and MTB trails, roads inside private and military areas, and narrow paths where it is unknown whether cycling is allowed. "Signed cycle paths" are cycle paths, roads with a cycling sign or roads on a cycle route such as the junction network.

Ridden or not. Every remaining road was sampled every 15 metres. A stretch counts as ridden when one of the rides passes within about 40 metres; short bits under 60 metres, such as junctions, take the value of their neighbours.

OpenStreetMap as of October 17, 2026. Strava rides fetched on October 17, 2026.

Never ridden
—
Unpaved in total
—
Still unexplored
—