What is healthchecks-mcp?

An MCP server for Healthchecks, the dead man's switch for cron jobs and scheduled tasks. Healthchecks works the other way round from an uptime monitor: instead of polling your service, it waits for your job to check in, and raises an alarm when the ping does not arrive. This server puts that data — and the ability to change it — inside an MCP client.

It speaks the Management API v3 and works the same against the hosted service and against a self-hosted instance.

Why

Because the interesting question is never "is anything red". You can see that on the dashboard. The interesting question is the next one: what did it print when it failed, when did it start, has it been flapping, and does the check even have the right schedule. Answering it means the ping list, then one ping's body, then the flip history — three lookups that a model can chain in one turn and a person does by clicking.

The second reason is quieter. Healthchecks has a handful of edges that are easy to get wrong and silent when you do: schedule discards timeout without saying so, tags are space-separated while keyword filters are comma-separated, channels replaces rather than merges, and a check created without channels alerts nobody at all. This server refuses the combinations that lose data and defaults the one that matters, instead of forwarding the mistake.

What it is not

• It is not a pinger. No tool here calls a ping URL. Pinging is a job's statement that it ran; a tool that could ping would let a model — or text a model happened to read — report success for something that never executed, and a monitoring system that can be talked into green is worse than no monitoring.
• It is not a Healthchecks administrator. Projects, members, ping keys and notification integrations are configured in the web UI; the Management API has no endpoints for them. list_integrations reads them so the write tools can refer to them by UUID.
• It is not a multi-project view. A Healthchecks API key belongs to exactly one project. To reach two, run two instances of this server with two keys.
• It is not an archive. The instance keeps 100 pings per check on a free plan and 1000 on a paid one, with no pagination and no way to reach past that, and flips only go back to the start of the month before last.