What you can do
Connect an assistant to mcp.canary9.com and it can work with your
monitoring the way you would:
- Create, update, pause, and resume monitored endpoints, and pull uptime and recent check results.
- Manage alert policies and review the alert event history.
- Create and manage notification integrations, from Slack and PagerDuty to email and plain webhooks.
Example usage
Suppose checkout latency looked bad overnight. In your assistant, you ask:
"Which of my endpoints failed overnight, and from which regions?"
The assistant pulls your alert events and recent check results, and reports that checkout.example.com alerted twice from eu-central-1 with response
times over your threshold. You follow up in the same conversation:
- "Add an HTTP check for the new api.example.com/v2/health endpoint every minute, and alert the on-call Slack channel." The assistant creates the endpoint and wires the alert policy to your Slack integration.
- "Mute the TLS expiry policy while we rotate certificates, and list everything currently muted." Done, with a summary of muted policies so nothing stays silenced by accident.
Setup is one command in most tools; the MCP server docs walk through Claude Code, Cursor, VS Code, and other clients.
Access your security team can reason about
Connecting an assistant uses an OAuth sign-in in your browser: no API keys pasted into config files, no shared credentials. Access is scoped to your organization and to the signed-in user, and you can revoke it at any time.
A dependable alert path
Checks, alert evaluation, and notification delivery run on deterministic rules, so the path from a failed check to your pager behaves exactly the same at 3am as it does in a demo. Assistants work on top of that foundation, and as we extend what AI can take off your plate, the alerting core stays predictable.
Why this approach
Assistants are genuinely good at the operational back-and-forth around monitoring, such as setting up checks, tuning policies, and summarizing incident history, so we made that first-class through an open protocol instead of building a proprietary chat window. Your monitoring meets your tools where they already are.