Opsgenie is shutting down: an alternative for AI agent alerts and approvals

Atlassian is retiring Opsgenie. New sales ended on 4 June 2025 and support ends on 5 April 2027, after which remaining data is deleted. If you migrate with Atlassian’s tool, Opsgenie is switched off 120 days after your migration date, which can be well before 2027.

If your AI agents send alerts through Opsgenie, or use it to reach a person before acting, you need a plan.

First, sort your agent alerts into two kinds

Agents tend to use an alerting tool for two different jobs:

1. Incidents. “The nightly sync failed.” “Error rate is 12%.” An engineer needs to investigate. Nothing is waiting on the answer.
2. Decisions. “Approve a refund of $1,240?” “Deploy to production now?” The agent is waiting and needs a yes or no.

Opsgenie handled the first well. The second was always a workaround: an alert someone acknowledged, with no structured answer going back to the agent.

Options for incidents

• Jira Service Management or Compass. Atlassian’s own path, with an automated migration tool. The least change if you are on Jira.
• PagerDuty. Mature schedules and escalation. Free for up to 5 users; Professional is $21 per user per month billed annually.
• incident.io On-call. Modern on-call with strong Slack-based incident response.

Options for decisions: redphonr

redphonr is built for the second kind. The agent calls an MCP tool and waits; redphonr reaches a person and returns their answer.

Opsgenie concept	redphonr equivalent
Alert	alert_human or escalate (no answer needed)
Acknowledge	request_approval returns approved or denied with a note
Escalation policy	Escalation flow: steps by channel or role, with delays
On-call schedule	Weekly on-call and backup rotation, with overrides
Notification rules	Reach me ladder per person: Slack, SMS, WhatsApp, Telegram, Discord, email, voice
Heartbeats	Use CronEcho for missed check-ins from jobs and agents

Migrating agent alerts in four steps

1. List every place an agent calls the Opsgenie API. Mark each as an incident or a decision.
2. Move incidents to Jira Service Management, PagerDuty or incident.io.
3. Move decisions to redphonr. Replace the alert call with request_approval, ask_human or collect_input, with a deny default. If the agent speaks MCP, add the server with claude mcp add --transport http redphonr https://mcp.redphonr.com/mcp.
4. Rebuild your rota as a redphonr escalation flow: Slack immediately, SMS after five minutes, a call to the on-call person after fifteen, then the backup.

# Before: an alert someone acknowledged, and the agent guessed what that meant
opsgenie.create_alert(message="Approve refund $1,240 for order 88213?")

# After: the agent waits for an actual decision
result = await session.call_tool("request_approval", {
    "message": "Refund $1,240 to order 88213? Stripe shows two captures.",
    "timeout_seconds": 1800,
    "default": "deny",
})

Timeline to keep in mind

Date	What happens
4 June 2025	Opsgenie no longer sold to new customers
Your migration date + 120 days	Opsgenie switched off if you used Atlassian’s migration tool
5 April 2027	Opsgenie support ends and remaining data is deleted

For a wider view, see 10 ways to alert a human from an AI agent and redphonr vs PagerDuty.