Closed beta — 10 design partners, free for 60 days

Your on-call shouldn't be
a human alert filter.

Saneops correlates and deduplicates alerts from Grafana, Datadog and PagerDuty, then writes the first draft of the RCA. Your L1 stops clicking acknowledge at 3 AM and starts fixing things that matter.

~80%
noise reduction
< 1 min
L1 triage time
100%
self-hosted
The problem

800 alerts a day. 90% are noise. The other 10% are five real incidents fired six times each.

If your team runs a 24/7 NOC or on-call rotation, you already know the pattern. Your L1 spends the night clicking acknowledge instead of fixing anything. Your seniors get paged for the same outage from four monitoring tools. By morning, nobody trusts the pager.

₹25 L
/ year of L1 work for a 4-engineer NOC team in India — most of it spent triaging duplicates.
6–8 min
average time an L1 spends per alert: read, look up runbook, decide, escalate or close.
3:14 AM
when your senior gets paged for an incident your L1 already triaged twenty minutes ago.
How it works

Three things, in this order.

1

Ingest

Point any alert source at Saneops via webhook. Grafana, Datadog, PagerDuty, Prometheus, custom JSON. No agents, no SDK, no data leaving your network.

2

Correlate & dedupe

Time-window grouping + content-hash dedup + CEL-based custom rules collapse 800 alerts into a handful of incidents. The first 50 lines of the runbook become unnecessary.

3

Explain & route

A first-draft RCA is generated for every incident — what changed, what's affected, what to check first. Then it's routed to the right channel via a visual workflow builder.

Features

Built by an SRE, for SREs.

Every feature exists because a real on-call engineer wished it did at 3 AM.

Alert correlation

Time-window + label-similarity grouping with tunable thresholds per tenant.

Content-hash dedup

Same payload arriving from four monitoring tools? One alert, four sources attached.

CEL-based drop rules

Filter known-noise alerts with the same expression language Kubernetes uses.

LLM root-cause analysis

First-draft RCA written for every incident. Bring your own Claude / OpenAI / Ollama key.

Visual workflow builder

8-tab workflow editor — overview, builder, canvas, YAML, inputs, secrets, versions, runs.

Encrypted secrets vault

Webhook tokens, API keys, SMTP passwords — encrypted at rest with your session key.

Works with

Plugs into the stack you already have.

No agents to install, no SDK to import. Webhook in, webhook out.

Grafana
Datadog
PagerDuty
Prometheus
Slack
Microsoft Teams
OpsGenie
Zenduty
Anthropic Claude
OpenAI
Ollama
Generic webhook
Your data, your infra

Self-hosted by default. Always.

Saneops runs as a Docker container on your own infrastructure. Alert payloads, incident history, runbook content — none of it leaves your network. Not even to us.

  • Single Docker image, ~108 MB compressed. No agents. No SDK.
  • Encrypted secrets at rest using your own SESSION_SECRET.
  • Anonymous telemetry beacon — opt out with one env var.
  • DPDP Act 2023 + GDPR-aligned. Customer data classification = your responsibility, our software follows.
deploy.sh
# 1. Load the image (we send you a tarball)
$ docker load -i saneops-beta-protected.tar.gz

# 2. Edit two values in docker-compose.yml
$ vim docker-compose.yml

# 3. Bring it up
$ docker compose up -d

✓ Saneops running on http://localhost:8000
Total time: ~10 minutes
Beta program

Free for 60 days. Ten companies. No catch.

I'm taking on ten design partners through the closed beta. You get the full product, free, self-hosted. In exchange: honest feedback, a 30-minute weekly sync, and a chance to shape what Saneops becomes.

You get
  • • Full product, no feature gates
  • • Direct line to the founder
  • • 50% off year 1 if you convert
We ask
  • • Real production deployment
  • • Bug reports + feedback
  • • 30-min weekly check-in
Terms
  • • 60-day free beta
  • • Self-hosted, your data
  • • Walk away anytime
Apply for the beta

Or email omkrjsr29@gmail.com directly.

About

Built by one engineer, in public.

Saneops is built by Omprakash Kumar from Jamshedpur, India. No VC, no marketing team — just one engineer who got tired of watching on-call rotations burn out good people.

The bet: most alert-fatigue isn't a tooling-volume problem. It's a correlation problem. Once a system can group, dedupe, and explain alerts the way a senior SRE does in their head, the L1 layer mostly disappears.

If that bet's wrong, the beta will tell us in 60 days. If it's right, Saneops becomes the layer between your monitoring and your humans.

FAQ

Questions we keep getting.

Is this open source?
No. Saneops is proprietary, source-available to paid customers under NDA. The beta ships as a compiled Docker image — you run the binary, not the source. This lets us stay a small focused team without the dynamics of a large open-source project.
Does any of my alert data leave my network?
No. Saneops is fully self-hosted. The only outbound traffic is (1) an anonymous telemetry beacon you can disable with one env var, and (2) calls to the LLM provider you configure (Claude / OpenAI / Ollama). Want zero outbound? Use Ollama and disable telemetry — the system works in a fully air-gapped network.
What does it cost after the beta?
Pricing finalises before the beta cohort converts. Beta participants get 50% off year 1 and locked-in pricing for two years. Walk-away with zero obligation if it doesn't work for you.
How is this different from Keep / FireHydrant / PagerDuty?
PagerDuty is on-call routing — it tells the right human. Saneops decides whether a human needs to be told at all. Keep is similar in scope but cloud-only and aimed at larger orgs. Saneops is built for the 50–500 engineer band where you can't justify a dedicated AIOps team yet.
What if I find a critical bug at 2 AM?
Hit the floating "💬 Feedback" button in the app — it pings me directly. Critical issues during business hours get a response in under an hour. Off-hours, same-day.
We're not in India. Can we still use it?
Yes. Saneops runs anywhere Docker runs. The beta agreement uses Indian governing law with optional MCIA arbitration for non-Indian customers. Most teams sign as-is.

Stop reading alerts. Start fixing things.

Ten beta slots. First-come, first-served.

Apply for the beta