trove-scot-mcp
MCP server for Scotland's historic environment, providing tools to search, filter, count, and locate heritage sites, listed buildings, scheduled monuments, and properties in care using Historic Environment Scotland's open ArcGIS API. Supports natural language queries for castles, monuments, archaeological sites, and more, returning locations, details, and trove.scot links.
README
trove.scot MCP Server
MCP server for Scotland's historic environment — 313K+ heritage records (castles, monuments, listed buildings, archaeological sites) via Historic Environment Scotland's open ArcGIS API, the backend for trove.scot. No auth required.
Tools
- search_heritage(term, sitetype?, council?, broadclass?, limit=50) — Search Scotland's National Record of the Historic Environment (Canmore). Filter by name, with optional site-type, council-area, and broad-classification filters. Returns site names, types, locations (lat/lon), and trove.scot links.
- get_heritage_by_id(canmore_id) — Full details of a specific site by its Canmore ID (the number at the end of a
trove.scot/place/{id}URL, e.g.52068for Edinburgh Castle). Returns the complete record with lat/lon and a trove.scot link. - count_heritage(term?, sitetype?, council?) — Count how many sites match a search without fetching the records. Useful to check whether a query is too broad (results cap at 1000) before searching.
- heritage_near(lat, lon, radius_km=1.0, term?, limit=50) — Find heritage sites near a geographic point (WGS84 lat/lon). Builds a bounding box around the point and returns sites sorted nearest-first, each with a
distance_kmfield. Optional name filter. Returns the{count, total_found, truncated, note?, sites}shape —total_foundcomes from a count query (the true number of matches in the area), andtruncated/noteflag when results were cut off by the 1000-per-query cap or bylimit. - search_listed_buildings(term?, category?, local_authority?, limit=50) — Search Scotland's 67K+ listed buildings. Filter by name/address, listing grade (
A,B, orC— A is highest significance), and local authority. Returns building name/address, category grade, date designated, location (lat/lon), and a link. - search_scheduled_monuments(term?, local_authority?, limit=50) — Search Scotland's scheduled monuments — nationally important archaeological sites protected by law. Filter by name and local authority. Returns monument name, class, area, location (lat/lon), and a link.
- list_properties_in_care(local_authority?, term?, limit=100) — List the castles, abbeys, standing stones, and other monuments that HES manages on behalf of the nation (e.g. Edinburgh Castle, Melrose Abbey, Callanish Stones). Filter by name or local authority.
Features
- No API key required — uses HES's free, open ArcGIS REST API
- Automatic retry with exponential backoff (the ArcGIS server can be slow under load — timeouts, 502s, and 503s are retried)
- British National Grid → WGS84 coordinate conversion built in — every result includes
lat/lon(the API only returns BNG eastings/northings) - Count-before-fetch with explicit truncation warnings — the API caps results with no pagination, so the tools tell you when a query was cut off and how to narrow it
- trove.scot deep links on every record (
https://www.trove.scot/place/{id})
Install
cd trove-scot-mcp
pip install -e .
Configure (Hermes)
Add to your Hermes config.yaml under mcp_servers:
mcp_servers:
trove-scot:
command: /mnt/HC_Volume_105667182/kimbo/.hermes/hermes-agent/venv/bin/python3
args: ["-m", "trove_scot_mcp.server"]
Or for Claude Desktop / other MCP clients:
{
"mcpServers": {
"trove-scot": {
"command": "python3",
"args": ["-m", "trove_scot_mcp.server"]
}
}
}
Development
pip install -e ".[dev]"
python -m pytest tests/ -v
API Notes
- Source: Historic Environment Scotland ArcGIS REST API at
https://inspire.hes.scot/arcgis/rest/services— the backend for trove.scot. No auth required. - Datasets: Canmore / National Record of the Historic Environment (313K+ sites), plus designation layers for listed buildings (67K+), scheduled monuments, and properties in care.
- Case handling: Text data in Canmore is stored UPPERCASE and the server's
LIKEis case-sensitive — but this server handles it for you (it wraps columns inUPPER()and uppercases your term). Just search normally, e.g.term="castle". Literal%and_in a search term are escaped (LIKE ... ESCAPE '\') so they match literally rather than acting as SQL wildcards. - Result caps (no pagination): The API silently caps a single query at 1000 (Canmore), 5000 (listed buildings), and 10000 (scheduled monuments). There is no way to page beyond this. The tools count first and set a
truncatedflag plus anotewhen a query exceeds the cap — narrow your query with filters (sitetype, council, category, local authority) for complete results. - Coordinates: The API returns British National Grid (OSGB36 / EPSG:27700) eastings/northings, not lat/lon. This server converts them to WGS84
lat/lonautomatically (~5 m accuracy — plenty for locating sites on a map). - Reliability: The ArcGIS server can be slow under load. The client uses a 20 s timeout and retries transient failures (timeouts, connection errors, HTTP 502/503) automatically with backoff (1 s, 2 s). If it reports the service as unavailable, wait a moment and try again.
- Data licence: Open Government Licence (per the API records).
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。