MCP Server
Your AI agent plans routes and reads the terrain
The MCP Server hands the Data API to Claude, Cursor and any MCP client as tools. It is remote and hosted, so there is nothing to install. Self-serve: create an account, add a payment method, and each tool call is billed as one Data API request.
Works with the tools you already use
Any MCP client
One server, the same answer in every client
Claude, Claude Code, Cursor and any client that speaks the protocol connect to one remote server. Tool names, argument schemas and descriptions come from the schema the Data API and the SDKs are generated from.
- Remote and hosted, with nothing to run locally
- Tools generated from the published schema
- Open-source Agent Skills for 30+ AI coding agents
~/.cursor/mcp.json
{
"mcpServers": {
"mountaya": {
"url": "https://mcp.mountaya.com/",
"headers": {
"X-API-Key": "pk_your_publishable_key",
"X-Secret-Key": "sk_your_secret_key"
}
}
}
}What your AI agent asks
Every Data API operation, as a tool
Your AI agent gets these tools the moment the server is connected.
Loops from a brief
Give a distance, an ascent and a start point, and get candidate loops back.
Route and re-route
Activity-aware directions on foot, on snow and by road bike, costed on slope, and on surface and way type on foot and by bike.
Analyze any line
Elevation profile and analysis of any line: ascent, descent, gradient, surface and travel time.
Read the ground
Elevation, slope and aspect at any point, or across an area to find high ground or a slope facing one way.
Reach zones and travel times
Isochrones from any point, plus time and distance matrices. Built for aid stations, marshal placement and coverage checks.
Find places
Geocoding, place search and outdoor points of interest near any point.
Name any position
Turn a GPS position into the nearest named place, with its distance.
Your itineraries
Read your organization's itineraries, drafts included, and get a short-lived link to download one as a route file.
AI agents and API
The questions your product asks every day
Ask in plain language through the MCP Server, or call the typed client. The Data API answers.
AI agent
Suggest three trail-running loops from Chamonix, around 14 km with 600 m of climbing.
{ activity: 'HIKING_AND_TRAIL', candidates: 3, distance: 14000, ascent: 600, start: [6.8712, 45.9231] }
3 loops · closest 13.8 km / +615 m
Three loops from Chamonix, each close to 14 km and +600 m. The tightest lands at 13.8 km with 615 m of climbing. Every candidate returns with full GeoJSON geometry, an elevation profile, and its surface breakdown.
Trace the fastest trail line from Chamonix to Argentière and tell me what's underfoot.
{ activity: 'HIKING_AND_TRAIL', waypoints: [[6.8712, 45.9231], [6.9271, 45.9822]] }
9.0 km · ~2h40 · asphalt → ground → gravel
A 9.0 km trail line up the valley to Argentière, about 2h40 on foot. It leaves town on asphalt, then runs on natural ground and gravel singletrack. Full GeoJSON geometry and the surface breakdown come back with it.
Break down the surface and gradient of this drawn segment before I send runners on it.
{ activity: 'HIKING_AND_TRAIL', coordinates: [[6.87126, 45.92307], [6.8713, 45.9231], …] }
+984 m · 19.9° max · asphalt → ground → compacted gravel
The segment climbs 984 m to a high point of 2,021 m with a 19.9° max gradient, a long, sustained pitch. Surface moves from asphalt to natural ground to compacted gravel, way type from street and footway, with steps, to path, then track.
Use case
Terrain and weather on one map, from Mountaya.
The depth of Mountaya's mapping ecosystem transforms the experience on our mobile apps. By integrating their contextual data, we give our users the terrain insight they need to anticipate and prepare their races in the best possible conditions.
Jonathan Levy

What will your AI agent plan first?
Create an account, then point Claude, ChatGPT or Cursor at the MCP Server and sign in. The docs walk you through each client.
Frequently asked questions
Paste mcp.mountaya.com into Claude, ChatGPT, Cursor or VS Code, sign in with your Mountaya account and pick a key. A client with no sign-in screen sends a key pair instead. The tool list, argument schemas and descriptions arrive from the server, so nothing is kept in sync by hand.
Any client that speaks the Model Context Protocol, including Claude, Claude Code and Cursor. To have a coding agent write your integration, the open-source Agent Skills teach the same APIs to 30+ AI coding agents.
One Data API request, billed at the Data API rate. The MCP Server needs no agreement of its own, and the pricing page lists every rate.
No. The server answers from Mountaya's own outdoor data and returns the result to your AI agent. Mountaya trains no model on what you send, and your AI provider's terms must exclude training on what it receives.
Yes. It runs on the same engine and the same credentials as the Data API. The default limits are 60 Data API requests a minute and 300,000 a month, and a signed agreement sets higher named limits.