trails-mcp
Enables searching outdoor trails by name or place, retrieving route details, elevation profiles, and optional weather forecasts via Windy.
README
trails-mcp
A multi-service MCP server for outdoor / mapping data. Service 1 wraps the open Waymarked Trails API (recreational routes from OpenStreetMap) and adds OpenStreetMap Nominatim geocoding so trails can be found by place name. Service 2 adds Windy point forecasts for trail weather, marine, and air-quality planning.
Trail search is read-only and needs no API keys. Windy forecasts require a
Point Forecast API key (WINDY_API_KEY).
What the AI can do
- Search trails by name or reference code (
GR20,E5, "Pennine Way") - Find trails near a place name in one step ("hiking trails near Zermatt")
- Find trails inside a lon/lat bounding box
- Get full route details: length, operator, description, website, Wikipedia, OSM tags, bbox, sub/super-routes, and (optionally) the full path geometry
- Get a route's elevation profile
- Get route geometry clipped to a map box (GeoJSON)
- Resolve a route's Wikipedia link
- Get a guidepost (signpost node)
- Get a route's waymarking symbol (SVG shield)
- Geocode any place name to coordinates
- Get weather/marine/air-quality forecasts at any lat/lon (Windy Point Forecast API)
- Browse all Windy models, parameters, and pressure levels with descriptions
All six Waymarked Trails flavours are supported via a flavour parameter on each tool:
hiking (default), cycling, mtb, riding, skating, slopes.
Coordinates: the upstream API speaks Web Mercator (EPSG:3857) internally. This server accepts and returns WGS84 lon/lat everywhere — the conversion is automatic. Route geometry is summarised by default (
geometry_detail: "summary") to keep responses small; pass"full"for every coordinate or"none"to omit geometry.
Tools
| Tool | Purpose |
|---|---|
trails_status |
Health/last-update of a flavour |
search_routes_by_name |
Fuzzy name/ref search |
find_routes_near_place |
Geocode a place + return nearby routes (primary entrypoint) |
find_routes_in_bbox |
Routes inside an explicit lon/lat box |
get_routes_by_ids |
Batch summary lookup by relation id |
get_route_details |
Full route detail (geometry trimmed by default) |
get_route_elevation |
Elevation profile |
get_route_segments |
Route geometry clipped to a box (GeoJSON) |
get_route_wikilink |
Wikipedia URL for a route |
get_guidepost |
Guidepost node detail |
get_route_symbol |
Waymarking symbol SVG |
geocode_place |
Place name → coordinates (Nominatim) |
get_point_forecast |
Weather/marine/AQ forecast at lat/lon (Windy; requires API key) |
list_forecast_options |
Catalog of Windy models, parameters, levels (no API call) |
Install & build
npm install
npm run build # compiles src/ -> dist/
npm test # unit tests for the geo/conversion helpers
Requires Node >= 18.18 (developed on Node 22).
Connect to an MCP client
The same stdio command works everywhere; only the config file differs. Run npm run build
first — the config points at dist/index.js.
Claude Code
Either run:
claude mcp add trails -- node /Users/chiragahuja/Desktop/trails-mcp/dist/index.js
…or add to ~/.claude.json (global) or a project .mcp.json:
{
"mcpServers": {
"trails": {
"command": "node",
"args": ["/Users/chiragahuja/Desktop/trails-mcp/dist/index.js"]
}
}
}
Cursor
Add to ~/.cursor/mcp.json (global) or <project>/.cursor/mcp.json (project-scoped):
{
"mcpServers": {
"trails": {
"command": "node",
"args": ["/Users/chiragahuja/Desktop/trails-mcp/dist/index.js"]
}
}
}
Live-development (no build step)
Point the client at the TypeScript source via tsx instead of the built file:
{
"mcpServers": {
"trails": {
"command": "npx",
"args": ["tsx", "/Users/chiragahuja/Desktop/trails-mcp/src/index.ts"]
}
}
}
Try it
Use the MCP Inspector:
npm run inspect
Or, in a connected client, prompt:
Find hiking trails near Zermatt and show details of the longest one.
…which chains find_routes_near_place → get_route_details.
Configuration (env vars)
Copy .env.example to .env.dev and set your Windy Point Forecast key (create one at
api.windy.com). For local dev, npm run dev loads
.env.dev automatically. For MCP clients, pass env vars in the server config:
{
"mcpServers": {
"trails": {
"command": "node",
"args": ["/path/to/trails-mcp/dist/index.js"],
"env": { "WINDY_API_KEY": "your_point_forecast_key" }
}
}
}
| Variable | Default | Purpose |
|---|---|---|
WINDY_API_KEY |
(unset) | Windy Point Forecast API key (Map/Webcam keys do not work) |
TRAILS_MCP_LOG_LEVEL |
info |
debug | info | warn | error (logs go to stderr) |
TRAILS_MCP_USER_AGENT |
trails-mcp/<version> (...) |
User-Agent sent to upstream APIs (Nominatim requires a descriptive one) |
Architecture
src/
core/ registry, tool type, http, errors, rate limiter, logger, result helpers
geo/ Mercator <-> lon/lat conversion, bbox helpers, geometry trimming
services/
index.ts registerAllServices() — the single place new services plug in
geocoding/ Nominatim client + geocode_place (shared, rate-limited)
waymarked/ Waymarked Trails client, response shapers, and one file per tool
windy/ Windy Point Forecast client, parameter catalog, forecast shaping
Adding another service = create src/services/<name>/ with a register() function and add
one call in src/services/index.ts. No existing tool files change. Tool names are namespaced
to avoid collisions, and the registry rejects duplicates.
Attribution & usage
- Trail data: © OpenStreetMap contributors, served by Waymarked Trails (Sarah Hoffmann), ODbL.
- Geocoding: OpenStreetMap Nominatim — used within its usage policy (max 1 request/sec, descriptive User-Agent), enforced in-process.
Please be considerate with request volume; these are free community services.
License
MIT (this wrapper). Upstream data/services keep their own licenses.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。