Your MCP server returns 200.
Is it actually alive to an agent?
AI agents reach your tools through the Model Context Protocol — and a generic uptime check can't tell whether your MCP server is truly up. Merlonix runs a real JSON-RPC initialize handshake and tools/liston a schedule, and alerts you when the server goes down, the handshake starts failing, latency regresses, or a tool is added, removed, or silently changes its input schema — the breaking change that quietly kills every agent calling it. It's the first monitor that speaks the protocol.
No credit card either way — start free, or trial the full workspace.
- Check cadence (Agency)
- 1 min
- Real MCP session
- Handshake
- Tool-schema alerts
- Drift
- One-shot checker
- $0
Where MCP servers break without anyone noticing
Three MCP failure modes a status-code check will never catch: a 200 over a broken handshake, a silent tool-schema change that breaks every agent, and a slow or auth-gated endpoint you hear about from a support ticket.
MCP servers are how agents call your tools and read your data. When one goes down, rejects the handshake, gets slow, or silently changes a tool contract, every workflow built on it breaks — and a generic HTTP uptime check never notices, because the URL still returns 200. Catching it takes a monitor that opens a real MCP session and compares the result to the last known-good one.
A 200 OK lies
Your MCP endpoint returns HTTP 200 while the protocol is completely broken
A generic uptime monitor GETs your MCP URL, sees a 200, and reports "up." But an MCP server can serve a 200 to a browser and still be dead to every agent: the JSON-RPC initialize handshake errors out, it advertises a protocol version your clients no longer accept, or tools/list throws. Claude, Cursor, and your customers' custom agents all fail — and your dashboard stays green because nothing spoke the protocol. The only way to know your MCP server is actually alive is to open a real MCP session and complete the handshake, which is exactly what a status-code check never does.
Silent tool-schema drift
You ship a tool-schema change and every agent that calls it starts failing — silently
The most dangerous MCP failure isn't downtime — it's drift. You rename a tool, tighten a required argument, or bump the input schema, deploy it, and the server stays perfectly "up." But every agent that was calling the old contract now sends arguments that no longer validate, and the calls fail downstream, inside someone else's agent, where you never see the error. A tool disappears in a refactor; a new tool ships without docs; the protocol version shifts under a dependency bump. None of it pages anyone, because none of it changes the HTTP status. By the time a customer reports "your MCP integration broke," it's been broken for days.
Latency + auth blind spots
The handshake gets slow, or auth starts rejecting, and you find out from a support ticket
MCP sessions start with a handshake round-trip; when that gets slow, every agent conversation built on your server feels sluggish to start, and there's no graph tracking it. Auth is worse: an expired token or a misconfigured gateway starts returning 401/403 to agents, which a naive monitor can't tell apart from "down" — so you either get no alert or a useless one. You need a monitor that distinguishes "up but auth-gated" from "actually down," tracks handshake latency over time, and can complete the full check behind your auth with a stored token.
How it works
MCP server monitoring that speaks the protocol: a real handshake and tools/list on every check, with alerts on downtime, latency, and silent tool + capability drift.
Merlonix opens a genuine Model Context Protocol session with your endpoint on a schedule, fingerprints what it exposes, and alerts you the moment the contract changes or the server stops responding — alongside the TLS, DNS, and uptime monitoring for the infrastructure underneath it, on one alert path.
01
Add your MCP endpoint URL — we open a real MCP session, not a status-code ping
Point Merlonix at your MCP endpoint (e.g. https://yourapp.com/mcp). On every check we POST a real JSON-RPC initialize handshake over the Streamable HTTP transport, read back the protocol version, server name and version, and advertised capabilities, then call tools/list (and resources/list and prompts/list when advertised). The result is a true up / degraded / down — a 200 with a broken handshake is correctly reported as down, because an agent can't use it. Try it first with the free one-shot checker; no signup.
02
Get alerted on tool + capability drift, not just downtime
On each run we fingerprint every tool's contract — its name, description, and input schema — plus the protocol version and server version. When the next run differs, you're alerted: a tool removed (agents calling it will break), a tool's input schema changed (the silent breaking change), a new tool added, a capability dropped, or the protocol version shifted. This is the signal no HTTP or generic uptime monitor can produce, and the reason drift stops surprising your customers.
03
Track handshake latency and distinguish auth-gated from down
Every check records the initialize round-trip latency, so a slow-handshake regression is a graph and an alert, not a support ticket. A 401/403 is reported as "up but auth-gated" rather than a false-positive outage — and you can store an access token on the monitor so Merlonix completes the full handshake and tools/list behind your auth, for a real health check of the surface your customers actually use.
04
Monitor MCP alongside the rest of your surface, on one alert path
Your MCP server rides on the same infrastructure as everything else — TLS certificates, DNS, the domain registration, uptime of the underlying host. Merlonix watches all of it in one account, so an MCP outage that's really an expired cert or a DNS change is correlated, not investigated three times. Alerts go to email, Slack, Teams, Discord, and webhooks — the channels your on-call already uses.
What a protocol-aware monitor actually gives you
Monitoring built for the people who ship MCP servers — where "up" has to mean the handshake completes, the tools list, and the contract your customers' agents depend on hasn't silently changed under them.
A one-time check tells you your MCP server works right now. Continuous, protocol-aware monitoring tells you the moment it stops — a failed initialize, a latency regression, or a tool whose schema changed in a deploy — before the agents (and customers) built on it start failing.
Real handshake
Every check is a genuine JSON-RPC initialize + tools/list over the Streamable HTTP transport — not an HTTP status-code ping. A 200 with a broken handshake is reported as down, because that's what it is to an agent.
Tool drift
We fingerprint every tool's name, description, and input schema each run and alert on any change — a removed tool, a silently changed schema, a new tool, a dropped capability, or a protocol-version shift. The failure mode no other monitor catches.
1 min
Check cadence on the Agency plan — your MCP endpoint is re-handshaked frequently enough that a downtime, a failed initialize, or a schema change is caught in minutes, not when a customer files a ticket.
First of its kind
No other uptime or monitoring tool checks MCP servers at all yet. Merlonix speaks the protocol, so you can be the team that already knows your MCP server is healthy before anyone asks.
Pricing
Flat monthly fee. Every MCP endpoint — prod, staging, per-region — and its TLS, DNS, and uptime included.
No per-endpoint charges. Pick the tier that fits how many MCP servers (and the infrastructure underneath them) you operate, and monitor them all from one dashboard on one alert path.
Starter
For a solo maintainer running a single MCP server who wants downtime, failed-handshake, and tool-drift alerts on one endpoint alongside its TLS and DNS.
$19/ month
- 15 monitored assets
- 3 seats
- 5 min check cadence
- SSL + DNS + vendor monitoring
- Email + Slack alerts
Team
For a dev-tool or API company operating several MCP servers (prod, staging, per-region) that need fast checks, latency tracking, and drift alerts on every tool contract they expose to agents.
$79/ month
- 60 monitored assets
- 10 seats
- 1 min check cadence
- SSL + DNS + vendor monitoring
- Email + Slack alerts
Agency
For a platform team or MCP host running many endpoints across customers or business units, with unlimited seats and one correlated alert path for MCP health, TLS, DNS, and uptime together.
$199/ month
- 250 monitored assets
- Unlimited seats
- 1 min check cadence
- SSL + DNS + vendor monitoring
- Email + Slack alerts
Compliance
For regulated-vertical teams that need continuous, audit-ready evidence.
$699/ month
- 500 monitored assets
- Unlimited seats
- 1 min check cadence
- SSL + DNS + vendor monitoring
- Email + Slack alerts
Beyond uptime
More than a 200 OK on every client site
The same account also watches the infrastructure and reputation layers a standard uptime check never reaches — each a live check you can turn on.
Know your MCP server is healthy before a customer's agent tells you it isn't.
Run the free one-shot check now — a real handshake against your endpoint in seconds — then turn on continuous monitoring with drift and downtime alerts. No credit card either way.