Grafana OnCall OSS Was Archived. Here Is How To Keep Phone Alerts Working.

If your Grafana OnCall escalation chain fires and nobody's phone rings, nothing is misconfigured. On 24 March 2026 Grafana archived OnCall OSS and disabled Cloud Connection — the hosted layer that delivered SMS, phone calls and push notifications to self-hosted installations. The scheduling, the rotas and the escalation logic all still work. The last mile is gone.

What actually changed

• 11 March 2025 — OnCall OSS enters maintenance mode. No new features, security fixes limited to CVSS 7.0 and above.
• 21 April 2025 — OnCall Cloud and Grafana Incident are merged into the unified Grafana Cloud IRM app.
• 24 March 2026 — the repository is archived into cold storage. Cloud Connection is switched off. The Grafana IRM mobile app stops supporting OSS stacks.

What survives: your deployment runs indefinitely, the code is still AGPLv3, and Grafana said plainly that it was not trying to force anyone onto its cloud and would support a community fork with best reasonable effort.

That deserves acknowledging. Grafana published the dates a year ahead, kept the licence open and told the truth about the trade-off. Compared with how Freshping ended — announced in January, dead in March, all data deleted by June — this is a good exit. It still leaves you owning a notification pipeline you did not previously own.

Option 1 — Wire up your own delivery

This is Grafana's own recommendation and the smallest change. OnCall already supports your own Twilio credentials for voice and SMS; Exotel and Zvonok work too. For push, route an outgoing webhook into Pushover, Gotify or ntfy.

Good if: your rotas and routing rules are elaborate and you do not want to rebuild them.

The honest cost: a Twilio account, a billing relationship, credentials in your config, and a delivery path that is now yours to monitor. There is a particular irony in needing a monitoring tool to tell you your monitoring tool's phone alerts have stopped. If you go this way, put a synthetic alert through the whole chain on a schedule and check it actually arrives — a silent paging pipeline is worse than no paging pipeline, because you think you are covered.

Option 2 — Grafana Cloud IRM

The managed path, with a migration route via Terraform or the OnCall API. Note that the Grafana Cloud Migration Assistant does not move OnCall or IRM resources, so this is a deliberate migration rather than a button.

The honest cost: IRM Pro is a $19/month platform fee including three active IRM users, then $20 per month per additional active user. For a four-person rota that is $39/month; for eight it is $119. If you self-hosted specifically to avoid per-seat pricing, this is the thing you were avoiding.

Option 3 — A different router

PagerDuty, incident.io, AlertOps and others all do alert routing properly, and if OnCall sat in front of a Prometheus stack you intend to keep, a router is genuinely what you need. Expect per-seat pricing here too.

Option 4 — Ask whether you needed a router at all

This is the question worth sitting with, because a good number of OnCall OSS installs were not really routing anything.

If your alerts overwhelmingly came from blackbox_exporter checking whether things respond, from certificate expiry, from a handful of HTTP probes — then you were running Prometheus, Alertmanager and OnCall to answer "is it up, and who do I wake". That is three moving parts and a Twilio account to do one job.

An uptime monitor with on-call built in does the whole chain: it runs the checks, decides something is down, and escalates. Nothing to self-host, no Twilio relationship, no pipeline to keep alive.

Monitorion is one of those. Being specific about what that means in practice: 26 monitor types, eleven alert channels including phone calls and SMS, multi-step escalation policies with on-call rotation, and checks from five global regions that must agree before you get woken — which kills the 3 AM false alarm from one flaky network path. Free covers 25 monitors with email alerts; phone, SMS and multi-region are on Pro at $29/month, and there is no per-seat charge at any tier, so the size of your rota does not change the bill.

What it is not: an alert router. It will not ingest your Prometheus alerts. If you need that, options 1 to 3 are your list and you should ignore option 4 entirely.

Choosing so this does not happen a third time

OnCall is the second alerting product to go cold in 2026, after Freshping. Three questions worth asking of whatever you pick next, including us:

• Is this the company's product, or one app inside a platform? OnCall was one app in a very large observability suite, which is exactly what makes a thing consolidatable.
• Can you export your configuration on the plan you are actually on? If export is a paid feature, you are one billing dispute from being stuck.
• Is any part of it yours? OnCall being AGPLv3 is why a fork is even possible. That is worth something real.

For what it is worth, our answers: monitoring is the whole company, CSV export is on every plan including free, and our private worker agent is open source. Hold every alternative to the same three.

See the full Grafana OnCall comparison · Start free — 25 monitors, no credit card