gachi-data
Provides Japanese station accessibility and hazard data via MCP tools, including toilet information, public toilet locations, and official hazard categories.
README
Gachi Data API — Japan Station & Accessibility Data (API · MCP · Open Datasets)
This is a hosted service. You do NOT self-host it. Get a free API key at https://api.gachi-tokusuru.com and connect to the remote endpoint below. The source in this repo is published for transparency; the data lives in the hosted backend, so a local clone will not return data.
Deep, obscure Japanese data you won't find anywhere else — stations, accessibility, vacancy, hazards. Hand-verified, English-first, built for AI agents. Available as an MCP server, a REST API, and free open datasets. One key works for both the API and MCP.
- 526 Tokyo stations — accessible / multipurpose toilets with floor, gender, equipment flags (wheelchair, ostomate, diaper table) and the nearest station exit (an original first-party value computed by spatial join — not in any raw dataset).
- 612 municipalities nationwide — public toilets with wheelchair / baby-seat / ostomate flags, address and coordinates.
- Free open datasets — Japan Station Master (entity-resolved, 9,145 stations nationwide) & Ridership 2000–2025, sharing one
station_id: https://github.com/eng213035/gachi-open-datasets (Zenodo DOI10.5281/zenodo.21199500).
Station names accept Japanese (新宿) or romaji (Shinjuku, Kita-Senju) for major stations.
Connect
- Endpoint:
https://api.gachi-tokusuru.com/mcp - Transport: Streamable HTTP (remote)
- Auth:
Authorization: Bearer <API_KEY>— free key at https://api.gachi-tokusuru.com
{
"mcpServers": {
"gachi-data": {
"url": "https://api.gachi-tokusuru.com/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}
Tools
| Tool | Argument | Returns |
|---|---|---|
get_toilet_by_station |
station (Japanese or romaji) |
Accessible toilets in a Tokyo station, with nearest_exit |
get_public_toilet_by_city |
city (Japanese) |
Public toilets in a municipality (top 50 for large cities) |
get_station_hazard |
station_name (Japanese or romaji) |
Official MLIT hazard categories at a station (flood / liquefaction / storm-surge), relayed live, cached 14 days — no derived score. Landslide & tsunami are license-restricted (available:false + link to official maps). Not a substitute for official hazard maps. |
REST endpoints
GET /v1/station-toilets/search?station=Shinjuku— accessible toilets in a Tokyo stationGET /v1/toilets/nearby?lat=&lng=&radius=&wheelchair=&ostomate=&diaper=— public toilets near a pointGET /v1/stations/{station_id}/hazard— official MLIT hazard categories at a station, relayed live (flood inundation-depth rank, liquefaction/landform, storm-surge presence). Values are returned verbatim from 国土交通省 不動産情報ライブラリ with attribution, cached 14 days — no derived score, and not a substitute for official hazard maps. Landslide & tsunami source layers are 一部非商用 (license-restricted), so they returnavailable:falsewith a link to the official hazard maps. Also available as the MCP toolget_station_hazard(station_name).station_idcomes from the Japan Station Master (e.g.st_00001); 423 of 425 stations have coordinates (2 remain unlocated →hazard: null).
All endpoints use the same Authorization: Bearer <key> and share one monthly quota. Full spec: /openapi.yaml, docs: /docs.
Example
curl -X POST https://api.gachi-tokusuru.com/mcp \
-H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
"params":{"name":"get_toilet_by_station","arguments":{"station":"Shinjuku"}}}'
Pricing
Free 1k · Pro $19/100k · All Access $49/200k · Business $149/500k. Full details: https://api.gachi-tokusuru.com
Paid plans are self-serve: after Stripe checkout the customer is redirected to /activate?session_id=…, which verifies payment, resolves the plan from the paid amount, and issues the API key on the page (idempotent per session — reload shows the same key). No manual key handling.
Operational notes (internal)
- Subscription cancellation is not yet wired to key revocation (no Stripe webhook in this build). A cancelled subscriber's key keeps working until manually disabled. Reconcile monthly: compare Stripe's active subscriptions against issued
key:*records and disable keys for lapsed subscribers. Acustomer.subscription.deletedwebhook → auto-revoke is the next phase. - Plan detection in
/activateis by paid amount (AMOUNT_TO_PLAN: $19/$49/$149). If a new plan reuses an existing amount, add an explicit mapping.
Licensing (two layers — read carefully)
- Code in this repository: MIT (see LICENSE). Applies to the server code only.
- Data returned by the API is NOT MIT. Each value carries its source in the response
attribution. It is derived from:- Accessible & public toilets — Tokyo Metropolitan Government, Bureau of Social Welfare & BODIK municipal open data (CC BY 4.0): https://portal.data.metro.tokyo.lg.jp/ · https://www.bodik.jp/
- Station names & train service status — ODPT, Public Transportation Open Data Center (CC BY 4.0): https://developer.odpt.org/
- Hazard categories, future population, land price — MLIT 不動産情報ライブラリ (Real Estate Information Library), official values relayed as-is per request (not a government-created dataset): https://www.reinfolib.mlit.go.jp/
- River flood forecasts & landslide alerts — JMA (Japan Meteorological Agency, 気象庁), relayed as published, not warnings issued by this service: https://www.jma.go.jp/bosai/
- Housing vacancy & municipality codes — Statistics Bureau of Japan (住宅・土地統計調査) via e-Stat & MIC (総務省): https://www.stat.go.jp/data/jyutaku/
- Nationwide stations & bus stops — MLIT 国土数値情報 N02/P11; English station names partly via Wikidata (CC0): https://nlftp.mlit.go.jp/ksj/ · https://www.wikidata.org/
- Derived by gachi-tokusuru.com (distinct from the official values above):
nearest_exit,nearest_station_km, and bus-stop counts are computed via spatial join. - Attribution is returned in every API response. Timeliness, accuracy and completeness are not guaranteed.
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。