I scanned every public MCP server. 91% were dead.

The scan

I pulled every public MCP endpoint I could find — MCP.so, Glama, PulseMCP, Smithery, the Official Registry, GitHub — and pinged each one the way an agent would.

Out of 2,181 remote servers, only 9% were actually alive. The other 91% were dead, broken-auth, or returning malformed JSON-RPC.

"Dead" isn't a TCP failure

A TCP ping is worthless here. What matters is whether the server speaks the protocol correctly.

I sent a real MCP initialize request over HTTP/SSE and validated the response shape against the spec. The top failure modes were:

• Connect timeout
• 401 with no Dynamic Client Registration support
• 200 OK, but malformed JSON-RPC body
• Tool schema drifted from what the registry advertised

Nobody is pinging these

Registry listings are static metadata. An author ships a server, submits to MCP.so, and moves on.

Six months later the container's gone, or OAuth broke, or the tool schema changed. The listing still says active. Nobody knows.

Who pays for this

• Indie authors lose users silently — the agent quits mid-call, the user blames the model, never comes back.
• Agent platforms shipping MCP support get blamed for ecosystem breakage they didn't cause.
• Internal teams can't monitor their own MCPs without a $400/mo Datadog SKU.

So I built AliveMCP

Every MCP in every public registry gets a free status page. What's actually running under the hood:

• Real initialize probe every 60 seconds — not a TCP ping
• 90-day uptime history and response-time line chart
• Schema-drift hash — alerts the moment your tool list changes
• Red flag when a server has been down over 15 minutes

Free public dashboard. Live at alivemcp.com.

The moat is the probe, not the dashboard

Any scraper can list 8,000 MCP servers. Two sites already do.

Nobody is speaking real MCP JSON-RPC against each one, hashing the tool schema every minute, and storing 90 days of response-time samples. That is what tells you why a user quit.

Pricing tracks who pays for what

• Public — free. Every listed MCP already has a /status page.
• Author $9/mo — claim your listing, custom alert webhooks, verified-author badge.
• Team $49/mo — private endpoints, Slack + email alerts, your own status subdomain.
• Enterprise $299/mo — SAML SSO, audit log, SLA PDF reports.

The cheap tier exists because I am an indie author too; $9 is below the threshold where anyone has to ask their manager.

If you ship an MCP server

Search your slug on alivemcp.com — it's probably already listed, and you probably want to know if it's green.

The waitlist sends you the first alert the moment v1 ships.

Not listed yet? Email your endpoint URL to hello@alivemcp.com — I'll add it today.