Pricing
One policy engine. Two things to govern.
Nirmata created Kyverno, the CNCF-graduated policy engine, and extended it to a second control point. Start with the infrastructure governance you already run. Add control over the AI coding agents your developers are already using.
- Govern your infrastructureKubernetes, cloud and infrastructure-as-code. Policy enforced in the pipeline, at admission and at runtime.
- Govern your AI coding agentsCost, productivity and security for Claude Code, Cursor, Copilot and Codex. Free 15-day trial for teams.
Kyverno runs in production at
- Bloomberg
- Coinbase
- Deutsche Telekom
- Spotify
- Vodafone
- Wayfair
Track one
Govern your infrastructure
Every plan runs the same Kyverno policies you already write. Start with a supported distribution; add fleet-wide control, remediation and compliance evidence as your environment grows.
- Nirmata Enterprise for Kyverno
Start here
Nirmata Enterprise for Kyverno
If you already run Kyverno OSS in production.
$1,233/month
$14,800 billed annually
- Same Kyverno, same APIs, same policies — no fork
- 0-CVE hardened images, rebuilt on disclosure
- 24/7 support from the Kyverno maintainers
- Backports on your pinned version and validated upgrade paths
- Nirmata Control Hub three tiers
Nirmata Essentials
When you need to see the whole fleet.
$1,800/month
20 clusters or 250 nodes included
- Centralized reporting across clusters
- Kyverno engine health
- Shift-left enforcement in the pipeline
- Assisted policy authoring
Add Nirmata Enterprise for Kyverno for $1,233/mo — or step up to Advanced, where it is included.
Most popular
Nirmata Advanced
For mission-critical production workloads.
$4,500/month
20 clusters or 250 nodes included
Everything in Essentials, plus Enterprise for Kyverno included
- Centralized policy sets
- Automated violation assignment
- Self-service exceptions
- Compliance automation and audit evidence
- Agent Hub, in Nirmata Control Hub — agents that find and fix Kubernetes issues for you
Nirmata Enterprise
Governance tailored to your environment.
Custom
Annual agreement
Everything in Advanced, plus
- Remediator Service Agent — opens PRs for Kyverno violations
- GitHub and GitLab integration
- Identity management (SCIM)
- Bring your own key for AI features
- Air-gapped and on-premises deployment
All plans include 20 clusters or 250 nodes. Additional clusters and nodes are priced on consumption tiers. Flexible support packages are available on any plan.
Kyverno OSS or Nirmata Enterprise for Kyverno
Same Kyverno. Same APIs. No fork. The difference is who carries the risk.
| Kyverno OSS | Nirmata Enterprise for Kyverno | |
|---|---|---|
| Is it the same Kyverno? | Yes — Apache 2.0, straight from upstream | Yes. Same upstream, same APIs, same policies — built and signed by the maintainers. |
| A critical CVE drops on a Friday night. Then what? | Wait for a community release, then rebuild, test and roll out yourself | Pull a 0-CVE hardened image. We rebuild on disclosure, under SLA. |
| Enforcement breaks in production at 2am. Who fixes it? | Your team, a GitHub issue and community goodwill | 24/7 enterprise support with a response SLA — from the maintainers. |
| How long can I stay on the version I am running? | Community-maintained across the release train | Backports on your pinned version, with validated upgrade paths. |
| Air-gapped or restricted environments? | Build and maintain it yourself | Supported builds, on the Enterprise tier. |
| Cost | Free | $14,800 per year |
Most teams move the first time a CVE scramble or a failed upgrade costs them more than a year of support.
Infrastructure-as-code
Scan and validate Terraform plans for misconfigurations before anything is provisioned.
- Advanced and up
Cloud
One set of policies across AWS, Azure and GCP, with a single reporting surface.
- Advanced and up
Track two
Govern your AI coding agents
AIControls gives engineering teams control over the cost, productivity and security of Claude Code, Cursor, Copilot and Codex from one control point. Start with one team on a free trial. Add governance for the whole organisation when you need rules for everyone.
Per developer who uses an AI coding agent. Bring your own model keys; we never bill for model usage.
Start free
Team
See, save and route, for one team of up to 50 developers.
$8/developer/month
Billed annually · $10 billed monthly
- Spend and budgets by developer, model and tool
- Route to any model on your own keys, including OpenRouter
- Open-weight models through 30+ providers
- Credit stacking — use your AI plan first, then your provider credits and API keys once developers reach its usage limit
- Cost per merged PR and team productivity
- Session activity and summaries, kept 30 days
- Secret blocking, DLP and prompt-injection detection
- MCP tool visibility and AI tool inventory
- GitHub or Google sign-in
Start 15-day free trial: TeamNo card required · 1–50 seats
Governance included
Business
Prove the return and set the rules, for an organisation of up to 500 developers.
$20/developer/month
Billed annually · $25 billed monthly · no minimum
Everything in Team, plus
- Org-wide productivity and an executive ROI report
- Routing policy and team model allowlists
- MCP governance — policy on every tool call, by identity
- Full session replay, revoke and search across teams
- SSO / OIDC and access bindings
- Session data kept one year
- Runs standalone or behind your LiteLLM or Envoy gateway
Add Advanced Security for +$6/developer/month, and Workloads per agent.
Enterprise
Everything included, on your infrastructure, for 500 developers and up.
Custom
Annual agreement · volume pricing
Everything in Business, plus
- Advanced Security included
- Workloads included, with shadow-agent detection in Kubernetes
- Nirmata cloud or self-hosted
- Contractual data residency
- Custom retention, with offload to S3
- SCIM and agent-registry integration
Advanced Security
+$6/developer/month
Detection and response for coding agents: behaviour baselines, session anomalies, attack sequences, risk scoring, findings with evidence, compliance mapping and SIEM export. An AI Judge settles borderline detections and an AI Analyst triages what is left.
Business add-on · included in Enterprise
Workloads
Per agent
Governance for autonomous agents running in Kubernetes: continuous discovery including shadow agents, admission policy, per-agent tool authorization and runtime signals.
Business add-on · included in Enterprise
US or EU hosting region on every plan. Regional Team pricing for businesses in India. Prices exclude taxes. Team is also available through GitHub Marketplace; Enterprise through AWS Marketplace.
What would this cost us?
Count the developers who use an AI coding agent, not your whole headcount.
Estimated per year, billed annually
$48,000
about $4,000 per month
200 developers × $20 Business
An estimate, not a quote. Model usage is billed by your providers, on your own keys.
For Business · 30 days
AI Engineering ROI Assessment
Buying for more than one team? Start with one pilot team of 20 to 50 developers, with no change to how anyone works. At the end you know what AI-assisted engineering is costing you, what it shipped and where it is exposed, in numbers you can take to finance.
What you get at day 30
- Cost per merged pull request, with the evidence chain behind it
- Adoption and effectiveness by team
- Avoidable spend, in dollars
- Seat rightsizing across Claude Code, Cursor, Copilot and Codex
- Risk findings — secret reads, write-capable MCP tools, unmanaged models
Three tools don't add up to one control point
General-purpose gateways do not measure engineering productivity. Productivity tools cannot route a model call or block a tool. Doing all three means buying three products and integrating them yourself.
Assembled in-house
A gateway, a productivity tool, a security tool, plus the glue
- Included: Spend visibility by model and provider
- Included: AI impact and ROI reporting
- Included: Model routing and failover
- Not included: One evidence chain from spend to shipped work
- Not included: Policy on every MCP tool call
- Not included: One audit trail across all three
- Not included: Three contracts, three renewals, your integration
AIControls Business
One control point, one contract
- Included: Spend visibility by model, provider, team and user
- Included: Cost per merged PR, with the evidence chain
- Included: Routing across providers, with credit stacking
- Included: Budgets that block, inline
- Included: Policy on every MCP tool call
- Included: One audit trail across cost, productivity and security
- Included: Built on Kyverno policy-as-code
At 300 developers an assembled stack runs roughly $10,700–13,700 per month in list prices, before security tooling and integration work. AIControls Business is $6,000 per month, or $7,800 with Advanced Security, billed annually.
Not sure which one you need?
They are bought by different people and solve different problems. Most companies start with one.
Start with infrastructure governance if…
You own clusters, and policy enforcement is already part of your job.
- Are you running Kyverno OSS and carrying CVE and upgrade risk yourself
- Do you need one report across every cluster for an auditor
- Are exceptions being approved in Slack threads
Start with AIControls if…
Your developers are using AI coding agents and nobody owns the bill or the blast radius.
- Is AI tooling spend rising faster than anyone can explain
- Can you say what any of it shipped
- Do you know which MCP tools your agents can write with
Questions people ask before they buy
I already run Kyverno OSS. What am I actually paying for?
The same Kyverno, plus 0-CVE hardened images rebuilt on disclosure, a 24/7 response SLA from the maintainers, and validated upgrade paths. A CVE disclosure or a breaking change becomes our problem rather than your on-call engineer's. The full comparison is above.
Can I buy Nirmata Enterprise for Kyverno on its own?
Yes — $1,233 per month ($14,800 billed annually), with no platform commitment. It is also included in Advanced and Enterprise.
Do I need both products?
No. They share a policy engine, not a purchase. Infrastructure governance is bought by platform and security teams who own clusters. AIControls is bought by engineering leaders who own an AI tooling budget. Running both means one policy language and one audit trail across clusters and agents — which is why existing customers tend to add the second one, not why anyone starts with both.
How does AIControls pricing work?
You pay per developer who uses an AI coding agent. Team is $8 and Business is $20 per developer per month, billed annually; paying monthly costs $10 and $25. Advanced Security adds $6 per developer on Business, and Workloads is priced per agent. Enterprise starts at 500 developers, includes both add-ons and is priced on an annual agreement with volume pricing.
What is credit stacking?
AIControls uses the AI plan you already pay for first. When a developer reaches the plan's usage limit, requests continue on your provider credits, such as startup credits from Anthropic or OpenAI, then on API keys shared across the team or another provider. You set the order. Your developers keep working, your credits get used before they expire, and pay-per-token spend stays down. It is part of routing and is included on every plan.
Do we need our own model keys?
Yes. You always bring your own provider keys, subscriptions or an OpenRouter key, and your providers bill you for usage directly. Nirmata never bills for model usage or marks it up. On Team, AI summaries and insights also run on your key.
What happens when a team passes 50 developers?
Team covers up to 50 developers in one workspace per company. Beyond that, or when you need SSO or rules that apply to everyone, you move to Business and keep your data and settings.
Does AIControls replace my LLM gateway?
It can, and it can also run behind one. On Business and Enterprise, if you already operate LiteLLM or Envoy, AIControls runs behind your gateway and you keep your routing layer. On any plan, an existing OpenRouter key works as one upstream.
Does "autonomous" mean I lose control of production?
No. Remediation is staged as a pull request. You review, verify and merge. A human stays on the loop by design, and every policy can run in observe mode first.
Which plans support restricted or self-hosted environments?
Enterprise, on both tracks. On the infrastructure side that covers air-gapped and on-premises deployment, including bring-your-own-key for AI features. AIControls Enterprise runs in Nirmata's cloud or self-hosted, with contractual data residency. Every AIControls plan offers a US or EU hosting region.
What happens after the 15-day trial?
Neither trial needs a card to start. AIControls Team continues when you add a card or subscribe through GitHub Marketplace. On the infrastructure side the trial does not convert on its own; if you want to continue, we send an annual agreement.
One policy engine, two control points, one audit trail.
Start a 15-day trial of either product, or put thirty days of real numbers behind your AI tooling spend. None of it needs a purchase order.