Pingoru now has an MCP endpoint. Ask Claude if OpenAI is down.

The thing I keep doing, twenty times a day, is tabbing out of Cursor to a status page. Is OpenAI degraded? Is the Stripe API actually down or is it just me? Cloudflare? Twilio? Each one is a separate tab, a separate URL I don't remember, a separate UI to scan.

So we shipped an MCP endpoint. Pingoru Pro accounts can now plug Pingoru into any AI assistant that speaks the Model Context Protocol: Claude Desktop, Claude Code, Cursor, Cline, Zed, Goose. Your assistant gets read-only access to every status page we track, which is currently 6,073 providers. You ask it "is OpenAI down right now?" and it answers from live data.

The actual tools

Ten of them. The names are dull on purpose, they're API endpoints, not marketing.

• list_providers searches the catalog by name or slug.
• get_provider returns the current state of one provider, including every component. AWS has 9,054 components in our database. Stripe has 47. The tool returns whatever's degraded at component-level, not just the rolled-up overall status.
• list_active_incidents returns everything currently firing across the catalog, with severity and start time.
• get_incident returns the full vendor timeline for one incident. If Twilio posted 17 updates over a 3-hour outage, you get all 17.
• provider_outage_history returns the past 90 days for any provider.
• search_components finds a specific service across the catalog. "Find me Stripe Checkout's status" works. So does "find me every us-east-1 service."
• list_monitors, get_monitor for what's on your own account.
• provider_uptime computes an uptime percentage over a chosen window.
• recent_incidents_global is the firehose: newest incidents across all 6,000-ish providers.

Full schema and example calls are at /support/mcp.

Why live data

Ask Claude "is OpenAI down right now?" without MCP, and you get a guess from training data that ended six months ago. The honest answer to that question changes every five minutes. Models can't know.

With Pingoru's MCP plugged in, Claude calls list_active_incidents, gets back JSON with the actual current state, and answers. We poll every five minutes, so the data is at worst a few minutes stale. Same data the dashboard runs on. No separate pipeline, no caching layer between MCP and the dashboard. If our database knows about it, MCP returns it.

What you set up

A single config block in your AI assistant's MCP settings. The full setup for Claude Desktop is:

{
  "mcpServers": {
    "pingoru": {
      "url": "https://pingoru.io/api/mcp/",
      "headers": {
        "Authorization": "Bearer pk_YOUR_API_KEY"
      }
    }
  }
}

You generate the API key at app.pingoru.io/api-keys. The full walkthrough for Cursor, Zed, and the rest is on the support page.

Plan gating and rate limits

MCP is on the Pro plan ($15/month). The three rate limit layers are:

1. 30 requests per minute, per key (burst)
2. 1,000 requests per month, per key
3. 5,000 requests per month, per account

These are conservative starter numbers. We'll relax them once we have real usage data. If you're a Pro account and hitting the cap legitimately, email us, we'll bump it manually.

What's not in v1

A few things I want to call out as deliberate omissions rather than oversights:

Write tools. Right now your assistant can read status but can't, for example, add a new monitor to your account ("hey Claude, start monitoring Twilio for me"). That needs a permission model I haven't designed yet, because the worst case ("Claude, delete every monitor I have") would be very bad. Read-only is the safe shipping point.

Webhook MCP. Stream incident events to the assistant instead of polling. The MCP spec supports it, our server doesn't yet.

Caching at the MCP layer. Every call hits the database. At our current scale that's fine. If it stops being fine, that's a good problem.

If you're using this and have ideas about any of those, find us on Bluesky. Or just hit reply on the next incident-alert email, those go to a human.

Pingoru independently monitors the status pages of every major cloud and SaaS provider, 6,073 of them as of this morning, and alerts you the moment one of your dependencies breaks. The Free plan covers 5 monitors. Sign up free.