Alert to all-clear, automated.
NOVA triages alerts, runs approved runbooks, and drafts stakeholder comms. Your engineers only take the pages that need judgment.
Plugs into the stack you already run
What NOVA takes off your plate
Six jobs your on-call does by hand today, done the same way every time, with the work shown.
Triage that groups the noise
Fifty pod alerts become one incident. NOVA correlates by service, deploy, and blast radius before anyone gets paged.
- checkout-api p95 above 1,280 ms
- pod checkout-7f9c restarted
- checkout-api 5xx burst
- payments queue depth rising
- synthetic: checkout flow failing
INC-2107 · checkout-api
Five alerts. One incident. One page sent.
Approvals where they matter
Read-only by default. Anything destructive waits for a tap in Slack, and every action lands in the audit log with who, what, and why.
policy: prod requires 1 approval
Context, assembled
Recent deploys, service owners, similar incidents: pinned to the incident before you open your laptop.
- deploy #482 · 14 min ago
- owner: payments-oncall
- similar: INC-2041, resolved
Runbooks that run themselves
Codify the fix once, in plain Markdown. NOVA executes step by step under guardrails and records everything it touches.
$ nova run rollback-checkout --incident INC-2107 guard deploy freeze on checkout-api ....... ok step roll back #482 to #481 .............. ok verify p95 < 400 ms ........................ ok INC-2107 resolved. Timeline attached.
Comms drafted before anyone asks
Status page updates and exec summaries, written from the live timeline. You edit, or just hit send.
draft · status page
14:32 UTC. Checkout latency is elevated for some customers. A rollback is underway; no data has been lost.
Postmortems with the timeline built in
Every alert, action, and approval is already on the clock. The review doc starts itself.
- 14:19first alert received, incident opened
- 14:23rollback approved by j.hartley
- 14:26p95 recovered, incident resolved
Rolled out in an afternoon
Connect
Point NOVA at PagerDuty, Datadog, Slack, and your clusters. Read-only until you say otherwise.
Codify
Import runbooks from plain Markdown, or fix something live and let the recorder draft the runbook while you type.
Hand off
NOVA takes first response on the alerts you choose. Start with one noisy service and widen as trust builds.
Run a drill. Watch it work.
Three recorded failures, replayed by the same engine that handles production incidents. Pick one and press run.
scenario
No drill running. Pick a scenario on the left and press “Run drill”.
Simulated drill data. Your incidents bring their own numbers.
The pager got quieter
“NOVA cleared a third of our pages before a human ever looked at them. The timelines it builds made our last postmortem the fastest we have ever shipped.”
“Status updates used to eat fifteen minutes of every incident. Now the draft is waiting, and support finally trusts what engineering publishes.”
“The guardrails are what sold our security team. NOVA proposes, we approve, and every action is in the audit log.”
Per engineer, nothing hidden
Solo
For the side project whose pager is also you.
$0
free forever
- 1 workspace, 1 service
- 3 runbooks
- 100 automated actions a month
- Slack and PagerDuty integrations
- Community support
Team
RecommendedFor on-call rotations that want their nights back.
$20per engineer / month
billed annually
- Unlimited runbooks and actions
- Approvals in Slack with full audit log
- Postmortem exports to Notion and Confluence
- 90-day incident timeline history
- Drill mode against real runbooks
- Priority support
Scale
For platform orgs with compliance in the room.
Custom
annual contracts
- SAML SSO and SCIM provisioning
- Private runners in your VPC
- Audit log export and retention controls
- 99.9% uptime SLA
- Dedicated Slack channel with our SREs
Questions engineers actually ask
What does NOVA actually execute, and what does it only suggest?
NOVA executes the runbooks you have approved, inside the guardrails you set. Anything marked destructive (rollbacks, restarts, scaling) waits for a human tap in Slack. Everything else it proposes with its reasoning attached, and an engineer decides. Every action, automated or approved, lands in the audit log.
Which tools does it integrate with today?
Paging: PagerDuty, Opsgenie, Grafana OnCall. Monitoring: Datadog, Grafana, Prometheus. Infra: Kubernetes, Terraform state (read-only). Comms and tracking: Slack, Jira, GitHub. If your stack speaks webhooks, the generic ingest covers the rest.
Do we have to rewrite our runbooks in a proprietary format?
No. NOVA imports plain Markdown runbooks as-is and asks clarifying questions about ambiguous steps. You can also fix something live while the runbook recorder watches, then commit the draft it produces. Runbooks stay in your repo, reviewed like any other code.
What happens when NOVA gets it wrong?
It fails loudly, not silently. Below its confidence threshold it pages a human instead of acting. Every automated change is reversible and logged with a one-command undo. And when an incident doesn't match anything it knows, it does the old-fashioned thing: wakes up your on-call with the context already gathered.
Where does our incident data live?
Encrypted at rest in the region you pick at setup (US or EU), encrypted in transit everywhere. Your logs and telemetry are never used to train models shared with other customers. On the Scale plan, runners execute inside your own VPC so commands never leave your network.
Can we start small before trusting it with production?
That's the intended path. Most teams connect read-only, let NOVA shadow their on-call for two weeks, and compare its proposed actions against what engineers actually did. Then they hand over one noisy, low-risk service and widen from there.
Put your on-call on rails
Connect your alerting in ten minutes. NOVA shadows your rotation before it ever acts.