nc-dmv
Provides live wait times and office directory for North Carolina DMV driver license and license plate agencies, sourced from the state's machine-readable feed.
README
@pipeworx/nc-dmv
North Carolina DMV MCP — live wait times at every NCDMV driver license office and license plate agency, plus the office directory the same feed carries. Keyless.
North Carolina and Oregon are the only two US states whose DMV publishes a machine-readable live wait feed. Every other state's is captcha-gated, blocked to cloud egress, broken, or only inside a phone app.
Tools
nc_dmv_wait_times(office?, service?, open_only?, sort?, limit?)— current average wait, longest wait, and open/closed status per office. 148 driver license offices;service: "vehicle"switches to the vehicle services / license plate agency units.nc_dmv_offices(office?, city?, zip?, service?, district?, limit?)— address, city, ZIP, phone, posted hours and NCDMV district. This feed is the only machine-readable NCDMV office directory that exists, so it doubles as the state's office locator.
Auth
None. Keyless and unauthenticated.
Data source
https://wait.services.ncdot.gov/IWT/IWTWeb/XmlStats.aspx?controller=IWT_UnitsListXML&UnitID=<n>
| UnitID | Root | Contents |
|---|---|---|
2 |
Driver Services | 148 driver license offices |
182 |
Vehicle Services | license plate agency units (1 as of 2026-07-29) |
Gotchas this pack handles for you
Closed-hours zeros are data, not gaps
Outside posted business hours every office reports OfficeStatus Closed with
AvgWT/MaxWT of 0:00:00. That is NCDMV reporting an idle queue — it is neither an
error nor a missing value, and it must not be read as "walk right in".
Responses therefore always carry offices_open, and when nothing is open the note says
the wait figures are idle-queue zeros. sort: "wait" puts open offices ahead of closed
ones for the same reason: a shortest-wait ranking would otherwise be won by 148 closed
offices tied at zero.
The feed is a tree, and the sentinel row lies about its parent
UnitID=2 returns 161 <Unit> blocks: one root (Driver Services, UnitLevel 1), eleven
districts (level 2), the 148 real offices (level 3), and a sentinel named Invalid. For
UnitID=182 the real unit sits at level 2 instead.
So the pack keeps leaves — any UnitId never named as another unit's ParentUnitId —
rather than filtering on UnitLevel, which survives NCDOT re-organising the tree.
The ordering matters: the Invalid sentinel hangs off a real office (Greensboro West,
ParentUnitId=143). Computing the parent set before dropping Invalid demotes that office
out of the results and yields 147 offices where NCDMV publishes 148. Invalid is dropped
first.
Addresses
-- is the feed's null. The separate <ZipCode> tag is empty on every single unit, so the
ZIP has to be parsed out of the one-line Address string
("521 S. Sandhills Blvd., Aberdeen, NC 28315"). That line is not consistently punctuated:
some offices embed newlines, omit the comma before the city, or drop NC altogether
("2479 E. Fifth St., Lumberton 28358"). The parser peels from the right — ZIP, then the
state token, then the city at the last comma — and returns the whole string in address
with city: null when the city cannot be separated from the street safely.
Current coverage: 118 of 148 offices carry an address at all (NCDMV leaves 30 blank);
of those, 118 ZIPs and 116 cities parse. offices_missing_address reports the gap.
Some entries are stale on NCDMV's side — Greensboro Express OLD still appears in the
feed. They are returned as published rather than second-guessed.
Wait units
AvgWT / MaxWT are H:MM:SS; the pack converts them to whole minutes.
Related
or_dmv_wait_times/or_dmv_offices— Oregon, the other live state wait feed.ca_dmv_offices— California field offices (no wait feed published).
Quick Start
Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
{
"mcpServers": {
"nc-dmv": {
"url": "https://gateway.pipeworx.io/nc-dmv/mcp"
}
}
}
Or connect to the full Pipeworx gateway for access to all 1392+ data sources:
{
"mcpServers": {
"pipeworx": {
"url": "https://gateway.pipeworx.io/mcp"
}
}
}
Using with ask_pipeworx
Instead of calling tools directly, you can ask questions in plain English:
ask_pipeworx({ question: "your question about North Carolina DMV data" })
The gateway picks the right tool and fills the arguments automatically.
More
License
MIT
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。