Pricing
Pay for workflow coverage, not mystery inference markup.
Kybernis Community is free and Apache-2.0. Managed plans charge for reusable skills, approvals, auditability, and support while bring-your-own-model usage stays your relationship with the underlying provider.
Typical request
first runMessage
“Can I book a demo next week?”
Qualify
Check calendar
Update CRM
Saved skill
Lead to Meeting
Next similar request already has a path.
Community
Kybernis Community
Self-host the Apache-2.0 core, bring your own model keys, and build on the same workflow and gateway foundation as the managed product.
Apache-2.0 community edition
OpenAI chat-completions and Anthropic messages adapters
Bring your own model keys and inference endpoints
Core planner, approvals, and workflow replay
Community docs, examples, and support
Managed cloud
Upgrade when oversight and operations start to matter.
Paid plans are for teams that want Kybernis to run as a production control plane: reusable skill coverage, approval discipline, channel operations, retention, environments, support, and procurement.
Managed plans
Start with flat monthly plans. Keep inference billing honest.
The managed product is priced around live approved skills, production channels, retention, environments, and support. If you bring your own model keys, Kybernis should not behave like a second token reseller.
For one team proving repeated work
Starter
$99
per month
Run one or two production workflows without taking a token tax on your own model keys.
Bring your own model keys with no token markup
Basic approval rules and run replay
14-day run, approval, and trace retention
One production environment
Email support
For teams turning reuse into operations
Growth
$499
per month
Expand across support, sales, onboarding, and internal ops with shared approvals and better operational visibility.
Bring your own model keys with no token markup
Shared approval inbox and reusable approval policies
90-day retention for runs, approvals, and replay
Staging plus production environments
Priority support
For multi-team and mid-market rollout
Scale
$1,499
per month
Operate Kybernis as shared production infrastructure with stronger approval control, audit history, longer retention, and faster support.
Bring your own model keys with no token markup
Approval queues, audit history, and rollout support
365-day retention for runs and approvals
Dev, staging, and production environments
Identity rollout planning and invoice billing support
For procurement, compliance, and private offers
Enterprise
Custom
annual or private offer
Package Kybernis for security review, procurement planning, private deployment scoping, and company-wide rollout.
Identity / SCIM design review and rollout planning
Private deployment scoping and architecture review
Custom retention planning and security review support
Marketplace and private-offer packaging support
Dedicated support and SLA scoping
How billing works
Use the platform fee for control-plane value. Keep provider cost separate.
This is the discipline that keeps Kybernis aligned with reuse instead of becoming another random bill on top of the model provider you already pay.
Bring your own models
The default Kybernis pricing model does not add a token markup when you connect your own Azure, OpenAI, Anthropic, Gemini, or compatible endpoint.
You keep the relationship with the underlying model provider
Kybernis charges for workflow coverage, approvals, audit history, and operational control
Best fit for teams that already manage their own inference spend
Managed inference when needed
If Kybernis fronts the model bill, provider spend should stay transparent and show up as a separate pass-through line item.
Platform subscription stays separate from provider usage
Useful for pilots, fast starts, or teams that do not want to manage keys yet
Keeps the invoice honest instead of hiding model cost in a mystery markup
Private bundles stay scoped
Outcome bundles, vertical deployments, and embedded partner deals should be quoted privately until unit economics and operational load are proven.
Use private pricing for tightly scoped workflows with known cost structure
Keep public pricing simple and horizontal
Avoid promising outcome margins before the delivery path is measured
Compare plans
The lower tiers prove value. The higher tiers buy oversight and procurement.
This is the line that shows up across strong infra and workflow products: core utility low, capacity in the middle, and compliance plus support higher up.
| Plan surface | Starter | Growth | Scale | Enterprise |
|---|---|---|---|---|
| Live approved skills | 5 | 20 | 75 | Custom |
| Production channels or surfaces | 2 | 8 | 20 | Custom |
| Seats | Unlimited | Unlimited | Unlimited | Unlimited |
| Bring your own model keys | Included | Included | Included | Included |
| Token markup on BYOM | None | None | None | None |
| Run and approval retention | 14 days | 90 days | 365 days | Custom |
| Environments | Production | Staging + production | Dev + staging + production | Custom |
| Shared approval inbox | No | Yes | Yes | Yes |
| Approval workflow support | Basic | Shared | Expanded | Custom |
| Audit history | Basic | Workspace-wide | Expanded | Custom |
| Enterprise identity rollout planning | No | No | Assisted | Custom |
| Directory sync / SCIM planning | No | No | On request | Scoped |
| Private offers and procurement support | No | No | Limited | Yes |
| Support | Priority | Faster priority | Dedicated |
Add-ons and programs
Charge separately for real delivery work instead of hiding it inside the subscription.
Onboarding, managed channel operations, and procurement-heavy deals are not imaginary. Price them honestly instead of hoping flat plans absorb every kind of delivery burden.
Managed execution plane and premium channels
Use this when you want Kybernis to help operate more of the delivery surface instead of only the workflow control plane.
Managed connector and channel operations
Provider-specific setup where message fees or support burden vary
Quoted separately so operations cost stays visible
Onboarding and first production rollout
Charge for the work that really exists: workflow design review, policy setup, QA, and launch support.
Best for teams shipping their first serious workflow set
Lets you recover implementation effort instead of hiding it in margin assumptions
Works especially well for partner or vertical deployments
Annual, private-offer, and marketplace packaging
Use annual commits and private offers for larger buyers instead of trying to stretch self-serve pricing to fit every deal.
Annual discounts for predictable rollouts
Private offers for AWS, Azure, and Google Cloud procurement paths
Custom bundles for embedded or regulated deployments
Startup and design-partner pricing
Good early programs reduce friction without poisoning the long-term price architecture.
Time-bound startup discounts
Design-partner rollout support tied to real feedback and proof points
Clear upgrade path once production usage is stable
Open-source and nonprofit pathway
The community edition should create goodwill and adoption, while managed plans stay available when operations and support matter.
Free Apache-2.0 community edition
Discounts for nonprofit, education, or ecosystem partners when appropriate
Managed cloud remains the paid upgrade for control-plane operations
FAQ
Pricing should feel like reusable work, not plumbing noise.
The public pricing model stays horizontal and legible. When a deal needs vertical packaging, private offers, or outcome bundles, it should be scoped and quoted honestly.
What is a live approved skill?
A live approved skill is a reusable workflow path your team has accepted for repeat production use. Drafts and experiments should not be billed like proven production skills.
Why do you not mark up bring-your-own-model usage?
If you already pay Azure, OpenAI, Anthropic, Gemini, or a compatible endpoint directly, Kybernis should not feel like a second tax on the same request. The platform fee pays for workflow reuse, approvals, auditability, and support.
What happens if Kybernis fronts the model bill?
Provider spend should stay transparent as a separate pass-through invoice line. The platform subscription and the underlying inference bill should not be mixed into mystery usage pricing.
What counts as a production channel or surface?
A production channel or surface is a live route where real work is being delivered or received, such as a support inbox, chat channel, webhook surface, or embedded product experience.
Is Kybernis Community actually open source?
Yes. Kybernis Community is the Apache-2.0 self-hosted core. Managed cloud, rollout planning, and private deployment support are paid upgrades.
Can I buy outcome bundles or vertical packaging?
Yes, but that should usually stay private and scoped. Outcome bundles make sense when the workflow, operational load, and provider cost profile are already understood well enough to quote responsibly.
Next step
What work should Kybernis learn first?
Start with one typical request. Let the agent do the work once. Approve the skill for next time.