AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻译准备中
官方 MCP 服务器,提供 Location Drive API——将 Claude、Cursor 或任何 MCP 客户端连接至 307.8 百万地点,覆盖 250 个国家和地区,254.2 百万分级建筑多边形,以及品牌情报。
AI 翻译
Official MCP server for the Location Drive API — connect Claude, Cursor, or any MCP client to 307.8M places in 250 countries and territories, 254.2M graded building polygons, and brand intelligence.
开发工具TypeScript · HTMLGitHub @Locationdrive在线运营者未确认
试用过了吗? 点一下就好,无需登录。
尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证正在加载完整 README。 以下为已保存的摘录。
Query the Location Drive location-data platform from Claude, Cursor, or any
MCP-compatible AI client: global POI search, graded building-footprint
polygons, brand intelligence, and premium data packs (EV charging, fuel
prices, full menus, multi-site hotel rates, multi-year review history) — using
your own Location Drive API key, so your plan, quotas, and usage dashboard
apply exactly as with direct API calls.
The dataset: 307.8M places across 250 countries and territories, 254.2M A–E
graded building polygons, 97 structured fields on every plan — Free included.查看完整 README ↗Query the Location Drive location-data platform from Claude, Cursor, or any MCP-compatible AI client: global POI search, graded building-footprint polygons, brand intelligence, and premium data packs (EV charging, fuel prices, full menus, multi-site hotel rates, multi-year review history) — using your own Location Drive API key, so your plan, quotas, and usage dashboard apply exactly as with direct API calls.
The dataset: 307.8M places across 250 countries and territories, 254.2M A–E graded building polygons, 97 structured fields on every plan — Free included.
| Tool | What it does | Cost |
|---|---|---|
search_places | Search the global POI database by name/keyword — ranked by prominence for general terms, by similarity for specific names | limit per page (default 20); max_results auto-pages and sums the cost |
find_nearby | Distance-sorted places around a coordinate (adds distance_m) | limit per page (default 20); max_results auto-pages |
brand_locations | Every location of a brand — one country or global — with the total | limit per page (default 20); max_results auto-pages |
get_place_details | Full record by ID; request any of the 97 fields via fields — premium packs (menus, hotel rates, EV connectors, fuel, review text) included on every plan; data_note when a requested field has no data for that place | 1 |
get_place_context | LLM-ready summary + structured data (sentiment keywords, not review text) for one place | 1 |
get_review_history | Merged multi-year review text history, duplicates removed, plus sentiment summary | 1 |
scan_reviews | Keyword scan of review text across up to 40 places near a point — matching snippets per place; neighborhood scale | the nearby search (its limit) + 1 per scanned place, reported as api_calls_counted |
get_hotel_rates | Compare room rates across booking sites for one hotel | 1 |
get_building_polygon | GeoJSON footprint with A–E accuracy grade | 1 |
find_building_at_point | Reverse lookup: which building contains this coordinate | 1 |
brand_footprint | Store counts, open-now, coverage, avg rating for a brand | 1 (summary) |
data_coverage | Per-country POI counts and polygon coverage stats | 1 |
All plans receive all 97 fields. Usage is counted per place requested: list
tools cost their limit (default 20); single-place and summary tools cost 1.
Free keys: 5,000 API calls/month, 20 results per call. A list request is
charged its limit whatever it returns, so the list tools never send more
than 20 unless you set limit; max_results (≤ 1,000) pages automatically
and returns the summed api_calls_counted.
Get an API key at locationdrive.com → Dashboard → API Keys.
Claude Desktop — add to claude_desktop_config.json:
{
"mcpServers": {
"locationdrive": {
"command": "npx",
"args": ["-y", "@locationdrive/mcp"],
"env": { "LOCATIONDRIVE_API_KEY": "ld_live_YOUR_KEY" }
}
}
}
Claude Code:
claude mcp add locationdrive -e LOCATIONDRIVE_API_KEY=ld_live_YOUR_KEY -- npx -y @locationdrive/mcp
Cursor / other MCP clients: command npx -y @locationdrive/mcp with the
LOCATIONDRIVE_API_KEY environment variable.
The same server runs hosted at https://mcp.locationdrive.com/mcp
(Streamable HTTP) — no install needed. Authenticate with your key as a
Bearer header.
Claude Code:
claude mcp add --transport http locationdrive https://mcp.locationdrive.com/mcp \
--header "Authorization: Bearer ld_live_YOUR_KEY"
Cursor — add to .cursor/mcp.json:
{
"mcpServers": {
"locationdrive": {
"url": "https://mcp.locationdrive.com/mcp",
"headers": { "Authorization": "Bearer ld_live_YOUR_KEY" }
}
}
}
Other clients: URL https://mcp.locationdrive.com/mcp with header
Authorization: Bearer ld_live_YOUR_KEY.
Ask your AI client things like:
reviews (current crawl),
historical_reviews (earlier crawls), and get_review_history (both merged,
de-duplicated). When a requested field comes back missing, get_place_details
returns a data_note: nothing was stripped, that place has no data for it.
get_place_context carries review_sentiments keyword counts, not review text.scan_reviews — it runs the nearby search, reads up to 40 places' review
histories (one API request each; ld_test_ keys are capped at 10 places),
and returns only the matching snippets.
Keyword matching is a coarse filter over user-submitted text: read the
snippets before drawing conclusions.limit per page (default 20, capped at the plan's results per request)
whatever the page returns, single-place and summary tools cost 1, and
scan_reviews costs the nearby search plus 1 per scanned place. Every tool
result includes api_calls_counted (from the response body, or the
X-API-Calls-Counted header when the body lacks it) — see
https://locationdrive.com/docs#api-call-counting.429 QUOTA_EXCEEDED means the key's monthly allowance is exhausted (the
refused request is not counted); it resets on the 1st at 00:00 UTC.ld_test_ keys work and return synthetic data without counting against
quotas — useful for trying the integration.LOCATIONDRIVE_TIMEOUT_MS overrides).npm install
npm run build # compiles src/ and typechecks api/
npm run inspect # MCP Inspector against the local build
LOCATIONDRIVE_API_KEY=ld_live_x node scripts/smoke.mjs # live API smoke test
MIT © Location Drive, LLC