roadsafety-mcp

An MCP service that lets AI assistants query Taiwan traffic accident data and accident hotspots (such as CBI). Public, read-only, no login required.

Unofficial: this service has no affiliation with the Ministry of Transportation and Communications (交通部) or the National Police Agency, Ministry of the Interior (警政署). It only reads public data on your behalf.

Ask your AI to install it with one sentence

Paste this sentence to the AI assistant you use (Claude, ChatGPT, Cursor and others). It finds the install method for your tool and asks you before running anything.

Please help me install the roadsafety MCP server at https://roadsafety.taux.io/mcp (Streamable HTTP, no login or token needed). Pick the right way to add it for the tool I am using; setup instructions are at https://roadsafety.taux.io/llms.txt. Tell me the command you plan to run first, and run it only after I agree.

Example questions

Connect

claude mcp add --transport http roadsafety https://roadsafety.taux.io/mcp

The URL is https://roadsafety.taux.io/mcp. Streamable HTTP, stateless, JSON-RPC batching is not supported.

Install in other AI tools

The URL is always https://roadsafety.taux.io/mcp; no token or headers are needed. Expand the tool you use:

Claude Code

Run this in a terminal:

claude mcp add --transport http roadsafety https://roadsafety.taux.io/mcp
Claude web and desktop (custom connector)
  1. Open Customize > Connectors.
  2. Click + Add, then choose Add custom connector.
  3. Enter any name, paste the URL below as the remote MCP server URL, and click Continue.
  4. For authentication choose No sign in (this service needs no login), then click Add.
https://roadsafety.taux.io/mcp
Cursor

Add the following to ~/.cursor/mcp.json (global) or the project's .cursor/mcp.json:

{
  "mcpServers": {
    "roadsafety": {
      "url": "https://roadsafety.taux.io/mcp"
    }
  }
}
VS Code (GitHub Copilot)

Add the following to mcp.json (the top level in VS Code's format is servers). You can also run MCP: Add Server from the Command Palette and follow the prompts:

{
  "servers": {
    "roadsafety": {
      "type": "http",
      "url": "https://roadsafety.taux.io/mcp"
    }
  }
}
Codex CLI

Add the following to ~/.codex/config.toml:

[mcp_servers.roadsafety]
url = "https://roadsafety.taux.io/mcp"
Gemini CLI

Run this in a terminal (by default it writes to the project's .gemini/settings.json):

gemini mcp add --transport http roadsafety https://roadsafety.taux.io/mcp
Clients that only support stdio (general method)

Add the following stdio entry to the client's MCP settings; mcp-remote bridges it to the remote URL:

{
  "mcpServers": {
    "roadsafety": {
      "command": "npx",
      "args": ["mcp-remote", "https://roadsafety.taux.io/mcp"]
    }
  }
}

Tools

Data sources

Most tools read data live from the Road Safety site (道安總動員, roadsafety.tw) run by the Ministry of Transportation and Communications, Department of Railways, Highways, and Road Safety (交通部路政及道安司). Accident records originate from the National Police Agency, Ministry of the Interior (警政署) and cause-of-death data from the Ministry of Health and Welfare (衛福部). get_location_npa_cases, get_township_drunk_ranking, get_county_drunk_enforcement and list_speed_enforcement_points use bundled public data from the National Police Agency and do not contact the upstream site.

Update time and provisional figures

Figures for the latest year are cumulative provisional values covering only some months, not a full year. Tools that return statistics (hotspot rankings, yearly history, accident locations, school areas, period comparison, improvement progress) include dataPeriod in their output, which states the data update date and the cumulative range. Rely on it, and do not read a drop in the latest year as a trend. The other tools (lists, hotspot cause breakdowns, bundled offline data) do not have this field.

Known limitations