An alert, not a question
The alert names the application, not a pod. The developer who owns it opens it themselves.
See how applications work
For SRE & platform teams
The Kubernetes SRE console that runs inside your cluster.
See what's broken, understand why, and fix it with approval-gated actions — real-time visibility, AI diagnosis, and an in-browser k9s terminal. Self-hosted: no SaaS backend, nothing leaves your network. One Helm command to install.
k9s terminal
kubectl in your browser
Prometheus + Grafana
plugs into your stack
Full management
workloads · Helm · security
Why this exists
Letting developers see their own logs means handing out kubeconfigs — and owning the scoping, the rotation, and the record of what was run. Most teams don’t. So the tickets come to you.
One platform where SREs and developers troubleshoot the same apps — with the access each role allows.
Admins keep the cluster and the troubleshooting. Developers read their own logs and app metrics, save what they need, and change nothing.
Both teams end up looking at the incident instead of at each other.
—
Usual Workflow
Logs without a ticket
They read the failure in their own service. REDIS_URL was never set. You are not in this conversation.
One incident, two owners
The cascade splits it in two. Upstream, a feature-store pod no node will take — yours. Downstream, the crash they are already reading, and a worker waiting behind it — theirs.
Pro · your own LLM key
The one that is yours
image-optimizer is OOMKilled, exit 137. You ask why. The answer names the limit it hit and proposes the patch that would raise it.
Pro
Nothing applies itself
You read the patch, then approve it. It is the only change to the cluster today, and it has your name on it.
See approval-gated writesWhat it recorded
Every write, every approval, every sensitive read. Hash-chained, kept a year. The developer’s morning is in there too, by name and source address.
See the audit trail
Developers watch their own apps — the logs, the metrics, how each release behaves.Now and then that surfaces the cause: a pod out of memory, a ConfigMap that never mounted.
You keep the cluster. All of it.One platform, two teams — each looking at the half that’s theirs.
Also in the box
- Cluster map
- Browser terminal with k9s
- Image scanning with Trivy
- Network policy view
- Node operations
- Resource Builder
The platform
Everything an SRE team needs, inside your own cluster.
Plugs into what you already run.
KubeManta runs inside your cluster and connects to the sign-in, paging, metrics and scanning you already have, so none of it has to leave.
KubeManta
in your clusterEnterprise
Pro
Free
Connect from outside, free on every plan
- Enterprise
- Pro
- Free
- MCP and the REST API are how agents and scripts reach KubeManta. Free on every plan.
- Compare plans
Bring your own model. Anthropic, OpenAI, or a local Ollama. Your key stays in the cluster, and with Ollama it runs fully air-gapped.
Pricing
Pay for people who can act. Everyone else watches free.
No node counting — ever. An admin is someone who can restart, scale, apply, run Helm, or open the terminal; viewers see everything and change nothing, and they're free and unlimited at every tier. Pro is $49 per admin seat per month; Enterprise is for SSO, AI governance, white-labeling, air-gap, and an SLA.
Free
$0/ month
one cluster · 1 admin seat · unlimited viewers
Triage, topology, terminal and workload actions on one cluster, plus one admin. No credit card, no trial clock.
Provided as-is; the Free feature set may change in future releases.
- One cluster · 1 admin seat · unlimited viewers
- Incident-first overview + failure cascade
- Read-only across every workload type
- Cluster resource map + network topology
- Metrics dashboards + resource trends
- Security posture — read trivy-operator reports
- Helm release viewing (values, history, diff)
- Browser terminal
- AI copilot — daily free quota
- K8s warning-event stream
- Alert rules engine — unlimited rules, Slack/Telegram/Email
- MCP server — read-only K8s tools for external AI agents
Enterprise
Custom
scoped to your org & requirements
For platform teams that need SSO, AI governance, white-labeling, air-gap, or an SLA. Tell us what you need — we'll scope a plan with you.
- Everything in Pro
- Assisted install & onboarding — set up with our team
- SSO — SAML, OIDC & LDAP
- Air-gapped deployment
- AI governance & data-residency controls
- Audit export & compliance support
- SLA + priority support
- Security-review assistance
Want to see it against your cluster shape before picking a tier? Book a live demo →
Questions people ask before buying
Short answers. If something here is unclear, email [email protected].
What counts as an admin seat?
An admin is anyone who can change the cluster — restart, scale, apply manifests, run Helm, or open the terminal. Viewers see everything and change nothing, and they are free and unlimited on every tier, including Free. You only pay for people who can act.
Do you charge per node, per pod, or per cluster size?
No. There is no node counting and no per-node overage — a 3-node cluster and a 300-node cluster cost the same. A subscription covers one cluster; the price scales with admin seats, not infrastructure.
What does the Free tier actually include?
Day-to-day operations on one cluster: incident triage, failure cascade, the resource map and network topology, built-in metrics, Helm and scan-report viewing, the browser terminal, and workload actions — restart, scale, delete — plus one admin seat and unlimited viewers. Alerting is included: unlimited rules delivered to Slack, Telegram or email. AI diagnosis, multi-cloud topology, long-retention Prometheus, and the alerting extras — auto-investigation of a firing, retained firing history, and the PagerDuty/Datadog/CloudWatch/webhook integrations — are Pro. No credit card, no trial clock, no countdown. It is meant to be genuinely usable, not a demo. The Free tier is provided as-is and free of charge, and what it includes may change in future releases — a change never reaches a version you have already installed.
Is there a free trial?
There is no separate self-serve trial, because the Free tier already runs the product in your own cluster, with no time limit on your license key. For paid plans there is a 14-day money-back guarantee from your first payment. If you need a time-boxed full-Pro evaluation for a team, contact us and we will issue one.
What happens if I add or remove seats mid-month?
You can change seat count at any time and your license stays valid throughout. Added seats are billed from your next billing period rather than charged mid-cycle; removed seats stop being billed from the next period. Nothing is pro-rated against you mid-month.
What happens when my subscription ends?
The software keeps running. It falls back to the Free tier feature set on that cluster — paid features re-lock, but nothing is deleted, disabled, or held hostage, and your data was never ours to hold: it lives in your cluster. There is no lockout, and nothing to uninstall.
Who processes payments, and can I get an invoice?
Paddle is the Merchant of Record. They handle payment, invoicing, and VAT/GST/sales-tax collection and remittance for your jurisdiction, so you get a proper tax-compliant invoice and we never touch your card details.
Does my cluster data leave my infrastructure?
No. KubeManta is self-hosted — it runs as two pods inside your own cluster, and we have no access to it. AI features are optional, use your own LLM key, and can be switched off entirely by an administrator; point them at a local model and there is no external egress at all.
Install
One Helm command, shaped to your cluster.
Answer three questions. The command below rewrites itself as you choose — copy it, run it on any Linux box with cluster access.
Where are you installing?
Who owns the load balancer?
How much do you want switched on?
helm install kubemanta \ oci://registry.kubemanta.com/kubemanta/charts/stable/kubemanta \ --version 1.2.0 \ --namespace kubemanta-system --create-namespace \ --set license.key="YOUR_LICENSE_KEY" \ --set global.hostname="your-domain.com" \ --set global.ingressClassName=alb \ --set global.tls=true \ --set rbac.allowWrites=true \ --set expertMode.enabled=true
Add --set license.key=YOUR_KEY — grab a free key below. Full flag reference at docs.kubemanta.com/getting-started/install.