Your site is down.Here’s why.

When your site, API or AI endpoint breaks, WhyDown works out why. Seconds after the alert, your team gets the cause, the fix and the evidence.

Start free See an example alert

Free for 20 monitors. No card needed.

Detect
Confirm
Investigate
Explain

14:05:58 UTC14:06:02 UTC14:06:14 UTC14:06:15 UTC14:06:43 UTC

api.acme.devUpFailingDown

HTTPS · checked every 30 seconds

20 min agonow

us-east212 msrefused
eu-west148 msrefused
ap-southeast264 msrefused

14:06:02First failed check, from us-east
14:06:14Confirmed: all three regions agree
14:06:43Cause and fix sent to #ops

One region failing never pages anyone.

All checks passing. Nothing to look into.us-east can't connect. Waiting for a second region to agree…Confirmed from three regions. Starting the investigation…

Investigating the outageCause found in 28 secondsread-only

DNSResolves to the same address as yesterday
CertificateValid for another 61 days
Port 443Connection refused from all three regions
ProvidersCloudflare reports no incidents
Recent changesDeploy a41f9c2 by sam, 4 minutes before the first failure

Recent changeConfidence high · 5 pieces of evidence

Deploy a41f9c2 at 14:02 broke the API. Four minutes later every region was refused on port 443, while DNS and the certificate stayed healthy.

Roll back a41f9c2.
Check what the deploy changed about the port the app listens on.

Cause and fix sent to #ops, 41 seconds after the first failed check

Read-only: it never changes your servers. Two regions must agree before anyone is paged.

The difference

Twenty tabs, or one message.

A typical alert

● DOWN checkout.acme.dev
Reason: SSL handshake failed

Then the search starts.

Server logs
DNS records
Certificates
Deploy history
Provider status pages
Team chat

A WhyDown alert

The TLS certificate for checkout.acme.dev expired at 03:12. It was issued on Jul 26 and never renewed.

Fix

Renew the certificate now.
Check that automatic renewal runs: the certificate for api.acme.dev was issued the same day.

DNS unchangedPort 443 openCertificate expired 03:12Issued Jul 26, never renewedNo provider incidents

Even when the answer is “it's not you.”

The investigator checks your stack and the services it depends on, and shows the check behind every claim. When the evidence doesn't point anywhere, it says it isn't sure.

Your change

A deploy at 14:02, then every region refused the connection.

Roll back a41f9c2, then check what it changed about the port the app listens on.

Confidence high · 5 pieces of evidence

Their outage

Vercel reports elevated errors in Washington, D.C. since 09:12.

Nothing on your side changed. Post a notice on your status page and wait it out.

Confidence medium · 3 pieces of evidence

False alarm

Ping fails, but HTTPS, TLS and DNS are healthy.

Your firewall drops ICMP. Switch this monitor to HTTPS, or close the incident and it won't count against your uptime.

Confidence high · 5 pieces of evidence

Your product runs on models and MCP servers. Watch those too.

An open port isn't proof. WhyDown asks the model for an answer and asks the MCP server for its tools, from every region.

LLM APIapi.anthropic.comevery minute

Sends a one-line promptPOST /v1/messages
Gets an answer back“OK” in 640 ms
Tells auth, rate limits and overload apart200

Works with Anthropic and any OpenAI-compatible endpoint, including your own gateway.

MCP servermcp.acme.dev/mcpevery minute

Completes the initialize handshake118 ms
Lists its tools14 tools
Still offers search_ordersfound

Catches the deploy that drops a tool your customers' agents depend on, not just the one that takes the server down.

Also monitors HTTP and JSON APIs, TLS certificates, DNS, TCP ports, ping, domain expiry and cron heartbeats.

Works with your agent

Ask your coding agent why it broke.

WhyDown has an MCP server. Claude Code, Cursor or VS Code can add monitors, read the diagnosis and pause checks for a deploy, without leaving the editor.

# Connect Claude Code
claude mcp add --transport http whydown \
  https://app.whydown.dev/mcp \
  --header "Authorization: Bearer wdk_…"

~/acme — claude

✻ Welcome to Claude Code!whydown MCP server connected

> Why did the API go down at 3am?

⏺whydown - list_incidents (MCP)(status: "resolved", limit: 5)

⎿Down 03:04–03:40 UTC · cause: provider outage

✻ Thinking… (esc to interrupt)

⏺The API was down from 03:04 to 03:40 UTC. Supabase had a database outage in us-east-1 for the same 36 minutes. Nothing on your side changed, so there's nothing to roll back.

> Pause API checks for tonight's deploy

⏺whydown - schedule_maintenance (MCP)(title: "API deploy", startsAt: "22:00Z", endsAt: "22:30Z", monitorIds: ["5f3c…"])

⎿Scheduled for api.acme.dev

✻ Thinking… (esc to interrupt)

⏺Done. api.acme.dev is in maintenance from 22:00 to 22:30 tonight, so nobody gets paged during the deploy.

>Why did the API go down at 3am?Pause API checks for tonight's deployTry "why did checkout go down?"

? for shortcuts

Free20 monitors every 3 minutes, 5 investigations a month

$29 / month for Pro100 monitors every 30 seconds, 300 investigations a month

The same product on both plansThe investigator, three regions, every alert channel, a status page and the MCP server. Pro adds monitors, faster checks, more investigations and your own status page domain. Compare plans

Know why beforeyour customers ask.

Set up your first monitor in under a minute.

Start free See pricing