outage-mcp
# outage-mcp
[](https://www.npmjs.com/package/outage-mcp)
[](LICENSE)
[](https://whatbroke.today)
Ask your AI if Cloudflare is down.
An MCP server that gives Claude, Cursor, and any MCP client live outage data from [whatbroke.today](https://whatbroke.today), which monitors 100+ service status pages continuously. No API keys, no signup.
```
You: Is GitHub down right now?
Claude: GitHub has 1 active incident: [MAJOR] Git operations degraded...
first seen 14:02 UTC. Details: whatbroke.today/incident/...
```
## Install
**Claude Desktop / Claude Code** — add to your MCP config:
```json
{
"mcpServers": {
"outage": {
"command": "npx",
"args": ["-y", "outage-mcp"]
}
}
}
```
**Claude Code one-liner:**
```bash
claude mcp add outage -- npx -y outage-mcp
```
Works the same in Cursor, Windsurf, and any MCP-capable client.
## Tools
| Tool | What it answers |
|------|-----------------|
| `whats_down` | "What's broken right now?" Active incidents first, then recently resolved. |
| `check_service` | "Is Stripe down?" Status and recent history for one service. |
| `get_incident` | Full detail for one incident, including AI root-cause analysis when available. |
| `outage_stats` | "What broke most this week?" Leaderboard and totals. |
| `list_services` | The 100+ services being monitored. |
## Notes
- Data comes from [whatbroke.today](https://whatbroke.today), which aggregates official status pages and adds AI summaries and RCAs.
- The upstream API is rate-limited to 5 requests/min per IP; this server caches responses for 60 seconds, which in practice keeps you under it.
- Severity `resolved` means the provider closed the incident. Everything else is treated as active.
## Development
```bash
git clone https://github.com/reshadat/outage-mcp
cd outage-mcp
npm install
npm run build
node dist/index.js # speaks MCP over stdio
```
MIT. Built by [Reshadat Ali](https://reshadat.com) between 10pm and 1am.
TDQS
Scored across 4 tools
Each tool targets a distinct query: check_service for a specific service's history, get_incident for incident details, list_services for the service catalog, and whats_down for current active outages. There is no confusion.
Three tools follow verb_noun pattern (check_service, get_incident, list_services), but 'whats_down' breaks the pattern with informal phrasing and no underscore.
4 tools is well-scoped for an outage monitoring server, covering the essential operations without unnecessary bloat.
The server covers the main use cases: listing services, checking specific history, current outages, and incident details. However, it lacks a global incident list or ability to search incidents across services, which is a minor gap.