Healthchecks Io MCP Server

@digitalronin/healthchecks-io-mcp

Unofficial, unaffiliated with Healthchecks.io. This is a third-party MCP server, not an official Healthchecks.io product.

An MCP (Model Context Protocol) server that lets an MCP client (e.g. Claude) manage Healthchecks.io checks via its Management API.

Status

v2 (read-only + mutating tools) published to npm as 2.0.0.

Setup

1. Get a Healthchecks.io API key from your account's Settings → API Access page. A read-only key only unlocks the read tools (see below). A read-write key unlocks everything, including deleting checks — as of v2, a leaked read-write key can silently disable or delete your monitoring, not just expose data. Use a dedicated key for this server and rotate it if you suspect it's been exposed.
2. Set it as an environment variable: HEALTHCHECKS_API_KEY.

Example MCP client config (stdio):

{
  "mcpServers": {
    "healthchecks-io": {
      "command": "npx",
      "args": ["@digitalronin/healthchecks-io-mcp"],
      "env": {
        "HEALTHCHECKS_API_KEY": "your-key-here"
      }
    }
  }
}

Scope

Read tools work with either key tier (except list_check_pings and list_integrations, which need a read-write key). Mutating tools always need a read-write key. delete_check and pause_check also require an explicit confirm: true argument — calling either without it returns an error describing what would happen, without making any API call. This is a safeguard against accidental single-call misuse; it does not (and cannot, over MCP's protocol) verify that a human — rather than the calling agent itself — actually approved the action.

Integrations (notification channels) have no create/update/delete API — Healthchecks.io only supports managing those through its web dashboard, so this server can't either.

The server detects your API key's tier once when it connects. On a read-only key, every write-tool call (and list_check_pings/ list_integrations) returns a clear error explaining the restriction instead of a raw API error or a missing tool.