HeyOnCall MCP Server

Introduction

The HeyOnCall MCP server provides a direct bridge between your AI assistant and HeyOnCall's incident alerting system. HeyOnCall is an all-in-one on-call platform that integrates on-call scheduling, alerting, website monitoring, and heartbeat monitoring into a single reliable product, serving DevOps, SRE, and engineering teams as well as solo developers.

This server exposes a focused capability: triggering critical, attention-demanding alerts that page the on-call user. Its Critical Alerts bypass Do-Not-Disturb and volume settings, are loud enough to wake someone up, and repeat until acknowledged, ensuring urgent issues are never missed.

With this MCP tool, an AI agent can autonomously escalate problems it detects—such as failed operations, service outages, or anomalies—by paging the responsible human with a concise message and relevant context.

Use Cases

• Escalate a detected outage or failure to a human. When an AI agent monitoring your infrastructure or logs discovers a critical problem it cannot resolve, it can page the on-call engineer immediately.

  • Sample prompt: I just checked our production API and it's returning 500 errors on every request. Page the on-call engineer with a message that the checkout API is down and include the dashboard URL https://status.example.com/api.

• Alert on a failed automated job or workflow. If a scheduled task, deployment, or data pipeline run by the agent fails, it can trigger a page so someone can intervene.

  • Sample prompt: The nightly database backup job failed with a disk-full error. Alert the on-call person that the Postgres backup did not complete tonight.

• Manual paging during an incident. A developer working with the assistant can ask it to page the on-call team directly when they've confirmed something is broken.

  • Sample prompt: We're seeing checkout failures spike on the dashboard. Page on-call now with the message "Checkout error rate above 40%, needs immediate investigation" and link https://grafana.example.com/checkout.

• Escalate security or anomaly findings. After an agent flags a suspicious event or threshold breach, it can raise a critical alert to ensure a human reviews it right away.

  • Sample prompt: Our SSL certificate for app.example.com expires in less than 24 hours and auto-renewal hasn't kicked in. Page the on-call engineer urgently about the expiring certificate.

• Trigger a wake-up alert for time-sensitive issues. For problems that genuinely can't wait, use HeyOnCall's Critical Alerts to bypass Do-Not-Disturb and reach the on-call person even overnight.

  • Sample prompt: The primary payment processor is unreachable and transactions are queuing up. Send a critical page to whoever is on-call so they wake up and fix this immediately.

• Notify on-call about degraded performance. When latency or health metrics cross a concerning threshold, the agent can page rather than let the alert go unnoticed.

  • Sample prompt: The API latency has jumped from 90ms to over 8 seconds for the last 10 minutes. Page the on-call engineer with a message about severe API latency degradation.