GaritaAI A CyberNest Secure product
Request a demo

Packages

Three packages. One governed gateway.

Priced to your environment — how many people use it, how many models are in play, and whether inference has to stay on hardware you own. Every package is quoted, so you are not paying for a tier shaped around somebody else’s organization. Connecting your own databases is a separate engagement that attaches to any of them — it is set out below.

Package

Garita AI Core

Governed AI access

For teams that need AI in production with real guardrails and a defensible record of use.

  • Policy-as-data guardrail engine
  • Prompt-injection and response scanning
  • Multi-provider routing with failover
  • Append-only audit trail
  • Per-user quotas and model access
Request a quote
Package

Garita AI Comply

Audit-ready governance

For regulated organizations that have to answer to auditors, customers and a security team.

  • Everything in Core
  • OWASP LLM Top 10 and MITRE ATLAS crosswalk
  • SIEM feed — Wazuh, Splunk, Sentinel, syslog
  • Evidence packets with SHA-256 manifests
  • Industry compliance profiles
  • vCISO-led policy tuning
Request a quote
Most protected

Garita AI Castle

Your data never leaves

For PHI, classified data, or export-controlled workloads where nothing may reach a third-party model.

  • Everything in Comply
  • On-premises model hosting
  • Regulated categories pinned to local inference
  • Private image generation
  • Air-gapped deployment
  • Dedicated engineering support
Request a quote

Add-on · any package

Connected data is its own engagement

Every other capability on this site is configuration. Reaching into your production databases is not — it is a project with a DBA, a data owner and, in regulated work, your counsel. Pricing it inside a tier would mean charging people who never connect a database for the work of the people who do.

What the engagement covers

Getting the connection safe before it is useful

  • A scoped read-only login, provisioned with your DBA — least privilege, verified before use
  • Schema review and catalog tuning, so the model is told about the right objects
  • Permit design — mapping who may reach which tables and categories onto real roles
  • Column masking decisions, made with the data owner rather than guessed
  • A pilot against your own questions before anyone depends on it
Compliance work it triggers

What has to be in place first

  • A BAA where PHI is in reach, and the risk assessment that goes with it
  • A documented lawful basis under GDPR, and applicable purpose-limitation, data-minimization and consumer-rights review under CCPA/CPRA
  • Sign-off from the data owner on every connection and every masked column
  • PHI stores pinned to local inference, which may mean standing up hardware
  • Evidence that the connection is read-only, kept current as grants change

It attaches to Core, Comply or Castle — a small team on Core can connect one database, and a Castle deployment can run without connecting any. The engagement is scoped per connection, so the second one costs less than the first.

Scope a data engagement

Side by side

What changes between packages

The boundaries are capability boundaries, not artificial gates — each tier adds the machinery the one below genuinely does not include.

CapabilityCoreComplyCastle
Guardrail engine and data classificationIncludedIncludedIncluded
Prompt-injection and response scanningIncludedIncludedIncluded
Multi-provider routing and failoverIncludedIncludedIncluded
Append-only audit trailIncludedIncludedIncluded
Token and budget quotasIncludedIncludedIncluded
Per-tenant strictness overrides—IncludedIncluded
OWASP / MITRE ATLAS crosswalk—IncludedIncluded
SIEM export—IncludedIncluded
Evidence packets with SHA-256 manifests—IncludedIncluded
Industry compliance profiles—IncludedIncluded
vCISO-led policy tuning—IncludedIncluded
On-premises model hosting——Included
Regulated categories pinned to local inference——Included
Primary-first standby failover——Included
Private image generation——Included
Air-gapped deployment——Included
Dedicated engineering support——Included
Connected data access (see below)Add-onAdd-onAdd-on

Quoting

What shapes the number

So the first call is a short one: these are the five things we ask about, and there are no surprises after them.

01

People

How many staff route through the gateway, and how many teams need separate policies.

02

Volume

Expected monthly tokens. Provider spend stays yours — bring your own keys.

03

Regulations

Which frameworks are in force, and whether any of them force inference on-premises.

04

Deployment

Hosted by us, in your cloud tenancy, on your own servers, or air-gapped.

05

Advisory

Whether you want vCISO-led policy tuning or your security team will manage it internally.

Not charged

What we don’t meter

Audit records, evidence packets and SIEM export are part of the product, not a data-retention upsell.

Questions

Pricing & deployment FAQ

Why are there no prices on this page?
Because the honest number depends on your deployment. A twenty-person clinic self-hosting one local model and a multi-tenant agency with an on-premises inference cluster are not the same product, and pretending otherwise means one of them subsidises the other.
Do we pay for model usage through you?
No. Bring your own provider keys and keep your own provider bill. Garita AI meters usage so that you can see and cap it, but the spend stays between you and the provider.
Is there a pilot?
Yes, and we recommend one. A scoped pilot against your own prompts and your own policy is the only way to know how the classifier behaves on your data before you commit.
Can we move between packages later?
Yes. The packages are the same product with different capability sets enabled, so moving up is configuration rather than a migration. Castle is the exception where it involves standing up inference hardware — that is a project, and we plan it with you.
Why is connecting a database priced separately?
Because it is the one capability that is not just configuration. Someone has to provision a scoped read-only login, review the schema, decide which columns are masked and map permits onto real roles — and in healthcare that sits behind a BAA and a risk assessment. Folding that into a tier would charge every customer for work only some of them need. It attaches to any package and is scoped per connection.
What is included in support?
Every package includes updates and email support during business hours. Comply adds vCISO-led policy tuning; Castle adds dedicated engineering support for the on-premises deployment.

Request a demo

See it run against your own policies

Thirty minutes. We switch on the regulations that apply to you, send the kind of prompts your staff actually send, and show you the audit record that comes out the other side.

Book 30 minutes directly

This form is sent to us directly and recorded as a sales enquiry — see the privacy notice for what we keep and for how long. Prefer to skip it? Write to info@cybernestsec.com.

Reach us directly

Emailinfo@cybernestsec.com Phone+1 (939) 355-4445 CalendarBook 30 minutes → Vendorcybernestsec.com →

Mon–Fri, 9 AM–6 PM AST. Existing customers with an incident: use the 24/7 line in your runbook.