Documentation

MCP server

Triage incidents and create monitors straight from Claude Code or Cursor.

The MCP server gives an assistant access to your monitors. Two scenarios justify it: triaging a failed job without opening the dashboard, and setting up a monitor for a cron job you just added without switching to the browser.

Setup

You need an API key from the workspace settings. A read-only key is enough for triage — with such a key the state-changing tools simply do not appear.

Claude Code or Cursor
{
  "mcpServers": {
    "tickwatch": {
      "command": "npx",
      "args": ["-y", "@tickwatch/mcp"],
      "env": {
        "TICKWATCH_TOKEN": "twk_…"
      }
    }
  }
}

What it can do

  • list_monitors, get_monitor — job states and schedule settings
  • get_runs — recent runs with exit codes and the output tail
  • list_incidents — what is broken now, or was before
  • get_workspace — plan, key scopes, how many monitors exist
  • create_monitor, pause_monitor, ack_incident — only with the write scope

What to keep in mind

A write key lets the assistant delete monitors and pause them. Start with a read-only key and issue a separate write key only when you actually need provisioning.

The server runs on top of the same public API and has no data access of its own. Everything it can do is bounded by the key’s scopes — revoking the key in the panel stops it immediately.

MCP server · Tickwatch