WhatsDownNow
Set up with your AI agent

Let your AI agent set up your outage alerts.

Paste one instruction into Claude, ChatGPT, Cursor, Grok Bot or any agent that can use MCP or call an API. It picks your services from 250 we track, creates the alerts and hands you a private link to manage them. No account, no API key.

1

Copy this into your agent

Fill in your services and where alerts should go. The instruction updates as you type.

Set up WhatsDownNow outage alerts for OpenAI, Stripe and Cloudflare, sent to me@example.com. Connect the MCP server at https://whatsdownnow.com/mcp and call create_alert_stack (or, without MCP, follow whatsdownnow.com/llms.txt and use the REST API). Then give me the private manage link it returns.

Asking a bot in a public post? Leave your email out and let the agent ask for it privately.

2

Your agent does the rest

About a minute, start to finish.

  1. Finds your services. Plain names work: "OpenAI", "Anthropic", "AWS", "Google Cloud", "Chase".
  2. Creates an alert stack with one create_alert_stack call. You're alerted when a service goes down, stays degraded for 15 minutes, or recovers.
  3. Gives you a private manage link, shown once. Open it to see your stack, send a test alert, reset the link or delete it. Keep it to yourself: anyone with it can change your alerts.
  4. You confirm your email with one click. Slack, Discord and webhook alerts work right away.

Connect the MCP server

Remote MCP over Streamable HTTP. It's stateless and needs no sign-in: https://whatsdownnow.com/mcp

ClaudeSettings › Connectors › Add custom connector, then paste https://whatsdownnow.com/mcp
ChatGPTTurn on Developer mode in settings, then add a custom connector with https://whatsdownnow.com/mcp and no authentication
CursorAdd to ~/.cursor/mcp.json (below)
Claude CodeRun the command below
VS CodeAdd to .vscode/mcp.json (below)
Any other agentIf it can read a web page and make HTTP requests, point it at https://whatsdownnow.com/llms.txt. It explains the REST API in plain text.
// Cursor: ~/.cursor/mcp.json
{ "mcpServers": { "whatsdownnow": { "url": "https://whatsdownnow.com/mcp" } } }
// Claude Desktop: claude_desktop_config.json (or Settings › Connectors › Add custom connector)
{ "mcpServers": { "whatsdownnow": { "url": "https://whatsdownnow.com/mcp" } } }
claude mcp add --transport http whatsdownnow https://whatsdownnow.com/mcp
// VS Code: .vscode/mcp.json
{ "servers": { "whatsdownnow": { "type": "http", "url": "https://whatsdownnow.com/mcp" } } }
create_alert_stackStart here: services (names or slugs) + email and/or webhook_url
get_alert_stackRead a stack: services, live status, channels, recent deliveries
update_alert_stackAdd or remove services, add an email or webhook
send_test_alertSend a sample alert to every active channel
rotate_alert_stack_keyIssue a new manage link; the old one stops working immediately
delete_alert_stackDelete a stack and stop its alerts
list_servicesFind a service by name or category
check_stack · get_status · list_incidentsLive status, before your agent runs a job
request_serviceAsk us to start monitoring a service we don't track yet
get_ai_fallbackHealthy alternatives when an AI API is down (or the whole /ai board)
report_issueReport a 5xx, timeout or rate limit your agent just hit

AI builder extras: the AI provider board, GET https://whatsdownnow.com/api/v1/ai/health, and probe history at GET https://whatsdownnow.com/api/v1/services/{slug}/history?days=7.

Request a service

Don't see something we should be checking? Request a service. Agents can call request_service or POST https://whatsdownnow.com/api/v1/service-requests.

No MCP? Use the REST API

One request creates the same stack. Names like openai are accepted. Full spec: OpenAPI 3.1.

curl -s https://whatsdownnow.com/api/v1/stacks -H 'content-type: application/json' \
  -d '{"name":"My stack","services":["openai","stripe","cloudflare"],"emails":["me@example.com"]}'

What your agent gets back

{
  "id": "stk_…",
  "token": "wdn_…",
  "manage_link": "https://whatsdownnow.com/stack#stk_….wdn_…",
  "IMPORTANT": "The token and manage_link are shown only once…",
  "next_steps": ["Tell the user to open the confirmation email sent to me@example.com and click the link…"],
  "confirmation_emails": [{ "email": "me@example.com", "confirmation_sent": true }],
  "services": [{ "slug": "chatgpt", "current": "up", "rules": { "on_down": true, "degraded_minutes": 15, "reports_per_hour": null } }, …]
}

Webhook stacks also return a webhook_signing_secret for verifying the X-WDN-Signature header. See webhooks.

Private by design

  • No accounts. We store only a hash of the management token. If a manage link may have leaked, open the stack and use Reset my link (or ask your agent to call rotate_alert_stack_key) while you still have it. If the link is lost entirely, it can't be recovered — ask your agent to make a new stack.
  • Email alerts start only after you click the confirmation link, and every email has one-click unsubscribe.
  • Up to 90 services per stack. Each network can create 5 stacks an hour.
  • Each service is checked about every 11 minutes against its official status page and our own checks. Alerts can lag or be wrong, so confirm critical decisions on the official status page.