Claude Code (MCP)

Pingniner runs an MCP server, which lets AI assistants read your monitoring data and manage monitors on your behalf. Instead of writing API calls you ask in plain language:

Which of my sites are down right now?

Add uptime monitoring for staging.example.com, checked every five minutes from London.

What was the latency on the marketing site during yesterday's incident?

It works with Claude Code and with any other client that speaks MCP.

Connecting

You need an API token first. When you create one, the connection command is shown underneath it with your token already filled in — copy that and skip this section.

Otherwise, run this in your terminal:

claude mcp add --transport http pingniner \
  https://app.pingniner.com/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

Then check it connected:

claude mcp list

The connection is available in every project on that machine. To connect from a different machine, create a second token rather than copying the first — that way you can revoke one without disturbing the other.

TIP

Give the token Read only if you want to ask questions without any risk of changes. You can always create a second token with Write later, or revoke and replace this one.

What it can do

Thirteen tools, in two groups.

Reading — available to any token with the Read ability:

Tool	Answers
account_status	What plan you are on, how many monitors you have used of your limit, which check intervals are available
list_monitors	Everything you monitor, filterable by type, status, group or tag
get_monitor	One monitor in full
monitor_history	Individual check results over a window
monitor_uptime	Uptime percentage for today, a week, a month or a year
list_alerts	Alerts raised, newest first — open ones only if you ask
server_metrics	The latest readings from a server's agent
list_locations	Where Pingniner can check from
list_groups	Your groups

Writing — needs a token with the Write ability:

Tool	Does
create_monitor	Adds any of the seven monitor types
update_monitor	Changes a monitor's settings
pause_monitor	Suppresses alerting for a monitor
resume_monitor	Turns alerting back on

A read-only token still connects — the writing tools simply refuse, and say why.

What it cannot do

It cannot delete anything. There is no delete tool and no plan to add one. Removing a monitor also destroys its history, so it stays a decision a person makes in the interface.

It cannot reach another account. The token fixes which account is in play, and nothing the assistant sends can change that. If you belong to two accounts, connect each with its own token.

It cannot change your billing, team or status pages. The tools cover monitors and the data they produce, nothing else.

It cannot invent check intervals. Your plan decides which are available, and asking for a one-minute check on a plan without one is refused — with the reason, so the assistant can tell you rather than silently doing something else.

Things worth knowing

Pausing does not stop checks

Asking to pause a monitor puts it into maintenance, which stops the alerting. Pingniner keeps checking and keeps recording history. Claude is told this plainly and should say so, but it is worth knowing yourself: if you want us to stop contacting a host entirely, pausing will not do it.

Servers need the agent installed

Asking Claude to add a server creates the record and returns an install key — the machine reports nothing until you install the agent on it. Claude is told to say this rather than reporting success, but the work is still yours to do.

Plan limits are enforced, and explained

At your monitor limit, creating another is refused with the actual numbers — using 30 of your 30 — rather than a bare failure. That is enough for the assistant to suggest removing something or upgrading, instead of trying again.

Rate limits

120 reads and 30 writes per minute, per token. Ordinary conversation stays well inside that; a loop that has gone wrong will not.

Being careful

An assistant with a Write token can create and change monitors on your account without asking you first. That is the point of it, and also the risk.

• Start with a Read token. Add Write when you actually want changes made.
• Watch what it does. Claude Code shows each tool call before running it.
• Check account_status occasionally, or the monitor list in the interface, to see that nothing unexpected has appeared.
• Revoke the token if a machine is lost, or if you stop using the connection.

Everything a token does is attributed to the person who created it, and each token shows when it was last used on the API tokens page.

Removing the connection

claude mcp remove pingniner

That detaches this machine. The token still exists and still works elsewhere — to stop it completely, revoke it under Account → API tokens.