Why your pager bill decides your reliability budget
Every engineering team buys three things: a way to page the right engineer, somewhere to run the incident, and a record of what happened afterwards.
That used to be one purchase. In 2026 it is usually three, and the newest part - an AI agent that proposes a root cause - is the one nobody prices.
The three pricing units in this market
Incident platforms bill per responder seat, per active user, or through a platform fee. AI SRE vendors bill per agent-second, agent-hour, or in credits: the layer that coordinates publishes prices, and the agentic layer mostly does not.
1. incident.io - the AI is on the plan, not beside it
incident.io runs the whole incident inside Slack or Teams: declare, assign roles, page, keep the status page current, then generate the post-mortem.
Team is $15 per user per month billed annually, on-call $10 more; Pro is $25 plus $20. A free Basic plan covers one team, and on-call can be bought standalone - see the pricing page shows.
Why it made the list: summaries and post-mortems come included, not as an add-on.
The exception is Investigations, the deeper agent that queries telemetry and opens fix pull requests - a paid add-on on Pro and Enterprise, unpriced.
2. Rootly - the modular price that quietly doubles
Rootly is the closest rival in shape: Slack, Teams or Google Chat native, with an agent that groups alerts and suggests shift cover.
It sells three products. Incident Response is $20 per user per month, On-Call another $20 - so an engineer who carries the pager and runs incidents costs $40, and the prices are additive, as Rootly's own pagestates. Under 100 employees you can save up to 50%, and under 25 employees pay what you can from $1.
Why it made the list: the startup tier is the most generous here, and the incident agent is included.
Rootly AI SRE - parallel investigation, confidence scores, evidence - is a third product sold alongside Incident Response, by quote.
3. PagerDuty - the incumbent with AI as a line item
PagerDuty remains the default paging layer in large organisations: deep escalation rules, hundreds of integrations, and an alerting history teams rarely migrate away from.
Incident Management is free for up to five users, then Professional at $21 per user per month billed annually ($25 monthly). Larger teams pay a platform fee per year - Starter $2,800, Essential $12,000, Plus $48,000 - plus seats (PagerDuty's pricing page.
Why it made the list: the most complete platform here, and the clearest case of AI arriving as an extra.
PagerDuty Advance and the SRE Agent are add-ons on Professional and Business, or included with a Reliability Platform plan. Neither is priced publicly.
4. Better Stack - on-call and AI on one invoice
Better Stack started as uptime monitoring and status pages, then added incidents and on-call. The status page is included, and free team members can read telemetry.
The on-call and incident plan is priced per responder licence: $34 per month, or $29 billed annually, with a $0 tier for personal projects and a $9 Slack-only option. See betterstack.com/pricing.
Why it made the list: readers are free, so the bill tracks the rotation, not the headcount.
The AI SRE agent - it investigates across logs, metrics, traces and errors, and can open a pull request - is bundled in, with model usage metered by consumption.
5. FireHydrant - capable, but the AI starts at Enterprise
FireHydrant is built around runbooks and automation logic: services, dependencies and severity rules are modelled explicitly, for teams that want a predictable response path.
It has the simplest price card here. Free covers ten responders, two runbooks and a status page; Pro is $25 per responder per month billed annually; Enterprise is quoted. On-call is billed by alert volume. See firehydrant.com/pricing.
Why it made the list: per-responder seats plus alert-volume on-call suits a small rotation.
The catch is the AI: summaries, transcripts, triage and retrospectives are all Enterprise-tier, so on Pro assume none.
6. Grafana Cloud IRM - per user, with token-billed investigations
Grafana approaches on-call from the observability side: if your dashboards and logs already live in Grafana Cloud, incident response is an extra module, not another vendor.
Incident Response and Management is free for three active users, then $20 per active IRM user inside a $19 monthly platform fee. AI is billed by consumption: $2 per million tokens beyond the allowance. See grafana.com/pricing.
Why it made the list: investigations are billed by the tokens they use, the fairest AI pricing here.
Grafana Assistant sits on a separate $20 per active AI user tier, so a heavy AI user who is also on call appears twice on the invoice.
7. ilert - European on-call with AI SRE still in beta
ilert is the strongest non-US option, popular where EU data handling matters. Alerting, escalation, call routing and status pages are one product, with an AI layer it calls ilert AI SRE.
Pricing is per user per month with a three-user minimum: Pro 19, Scale 39 and Enterprise from 49 billed annually, plus $10 for call routing. Those figures print without a currency, so confirm dollars or euros. See ilert.com/pricing.
Why it made the list: it names its AI precisely, and admits it is not generally available.
AI grouping, semantic triage and postmortems are labelled closed beta, and each plan carries a monthly AI credit allowance - 250 on Pro, 1,000 on Scale, 5,000 on Enterprise.
8. Datadog - the option you already pay, priced per seat
Datadog sells on-call, incidents and the Bits agents beside the monitoring most teams already buy: one less integration, but the workflow locks to one vendor.
On-Call is $20 per seat per month billed annually and Incident Management is $30, or $40 for the bundle. AI agents are metered in AI Credits - $500 for 500 credits a month, or $1.30 on demand - per Datadog's price list.
Why it made the list: Bits Investigation can investigate and remediate autonomously.
That autonomy is why the meter deserves a second look: a vendor that can fix production itself will spend your credits doing it, so set a cap first.
The AI SRE layer that publishes no price
Three vendors sell the same promise - agents that go on call so your engineers do not have to - and none publishes a number. Resolve AI sends you to a pricing form, Traversal has no pricing page at all.
It is the same inside the platforms you can price: Rootly AI SRE, incident.io Investigations and Splunk's AI SRE are quote-only, and Mezmo folds root-cause analysis into a licence quoted by data volume. See Resolve AI, Traversal and Mezmo.
The metered alternative: pay per agent-second, not per seat
AWS prices its DevOps Agent at $0.0083 per agent-second - thirty cents for a ten-minute investigation - with nothing charged while it is idle (aws.amazon.com). Azure bills its SRE Agent in Azure Agent Units - a fixed always-on charge plus token-metered work - on azure.microsoft.com.
Among the AI-native vendors, Cleric publishes real numbers: Team $700 a month for 1,400 credits, Pro $1,500 for 3,000 and Scale $3,000 for 6,000, at fifty cents a credit, with its unit economics printed (cleric.ai/pricing). Edge Delta starts at $20 a month including $20 of credits, with a spend cap built in (edgedelta.com).
Two things to check before you sign anything
Opsgenie is closing: new sign-ups ended in June 2025 and support ends on 5 April 2027, after which access is cut off and data deleted unless you migrate to Jira Service Management.
Squadcast no longer exists under that name: it is now SolarWinds Incident Response, its old pricing address returns a 404, and its $15 per user starting point sits in SolarWinds' list.
How to choose without a three-week trial
Start with the rotation, not the product. Count the engineers who will accept a page: that number times the per-responder rate is your real comparison - and why a $40 tool can beat a $21 one.
Finally, treat the free tiers as the trial: incident.io, Rootly, PagerDuty, FireHydrant, Better Stack, Grafana and ilert all let one team run free.
The Honest Takeaway
The boring half of this market has become transparent: eight platforms publish per-responder or per-user prices and most give you a free tier. The interesting half has not - the vendors selling autonomous investigation quote privately or meter you in credits or agent-seconds.
That is not a scandal - a vendor buying inference to keep your service up cannot charge a flat $20 a seat forever. It does change which line of the budget is worth arguing about, though.
Cover photograph: Manmakro, CC BY-SA 4.0, via Wikimedia Commons.