Skip to content
Zenith
Self-hosted data plane · Managed policy control plane

Control AI inference inside your network.

Buy Enterprise, download a signed .lic certificate, install the gateway, and activate against the issuer. Point any OpenAI-compatible client at your gateway — no proprietary SDK.

Enterprise license · ₺54900 per year

OpenAI
Compatible
Use standard SDKs and curl against gateway /v1. No proprietary client required.
Your
Network
Prompt and response bodies stay on the customer gateway path to upstream providers.
No
SDK lock-in
Zenith is a control plane — not an agent runtime or proprietary SDK.
Issuer
Reachable
Activation and managed policy sync need issuer reachability. Hard air-gap is not claimed.
OpenAI compatible

Keep your client. Change three values.

Use the official OpenAI SDK or any compatible client. Set the base URL, bearer key, and model for the Zenith surface you chose.

Cloud

Hosted, chat-first API

SDK guide →

Use your zk_ key from the dashboard with models, chat completions, and responses.

openai.py
base_url = "https://zenithgw.com/api/v1"
api_key = "zk_…"
model = "auto"
View Cloud API reference →
Enterprise Gateway

Your network, full /v1 surface

SDK guide →

Use a zgw_ key and make your configured pool alias the model value.

openai.py
base_url = "https://your-gateway-instance.example/v1"
api_key = "zgw_…"
model = "your-pool"
See Gateway API scope →

Inference stays on your path

Zenith Cloud issues licenses and syncs policy metadata. Your gateway routes inference to providers — Cloud is not an inference body store.

Client
OpenAI SDK / curl

Any OpenAI-compatible client points at your gateway /v1 endpoint.

Your network
Zenith Gateway

Admission, pools, routing, failover, and encrypted provider keys with persistent master.key storage.

Upstream
Providers

OpenAI, Anthropic, Gemini, Cloudflare, and OpenAI-compatible endpoints.

Control plane
Zenith Cloud

License activation, managed policy snapshots, staged rollout — not prompt/response storage.

Why teams buy Zenith Enterprise

Capability claims — not a published uptime SLA.

Routing & failover

Pool-based routing with health-aware selection and automatic next-healthy failover when a provider degrades.

Governance

Managed policy sync, admission controls, and operator-visible policy status — sync errors alone do not fail /readyz.

Config promotion

Validate, diff, apply, and roll back versioned config bundles. Credentials never appear in exports; rollback appends a forward revision.

Observability

Live /readyz and /metrics are the source of truth. Analytics and alerts use metadata — not prompt bodies.

See Smart Routing in Action

When a provider fails, Zenith reroutes automatically. Watch the direct and controlled paths side by side.

Direct API Call
OpenAI · gpt-4o · manual retry
> live routing trace
OpenAI · gpt-4o
Live
Zenith Smart Router
Auto · Multi-provider · automatic failover
> live routing trace
Auto · Multi-provider
Live
Live enterprise environment

Would you like to see the Enterprise demo?

Explore governance, scoped service accounts, managed gateway policy, routing analytics, and deterministic AI scenarios in a resettable environment.

Real Zenith data-plane and policy checks
No credentials or API keys exposed
Two-hour verified demo session
Environment resets nightly

Verify your email

We will send a six-digit, one-time code. Your email is used only to authorize this demo session.

Rate-limited. Demo credentials and API keys are never sent to the browser.

Operational proofs

What operators verify in a pilot — not synthetic latency theater.

/readyz

Structured readiness for Postgres, governance, and master key — live deployment truth.

Policy status

Managed / local / expiry / sync error visible in Settings and GET /admin/api/policy-status.

Secret-free exports

Config, providers, and edge bundles redact live API keys. Provider row deletes are skipped on apply.

master.key volume

Encrypted secrets need Postgres + master.key restored together. Persist the volume across recreates.

Enterprise Gateway license

Annual license. Self-hosted inference. Assisted activation against the issuer.

Self-hosted

Enterprise

Self-hosted Zenith Gateway with perpetual license. Inference stays on your network; minimal issuer control-plane egress.

₺54900
per year
  • Yearly license
  • Self-hosted on your infrastructure
  • Unlimited inference
  • Post-quantum encryption
  • Zero vendor telemetry mode (ZENITH_TELEMETRY=off)
  • Prompts/responses never sent to Zenith Cloud
  • Priority security patches
  • Dedicated onboarding

After payment verification, your signed .lic appears on Licenses for download.

Not ready for self-hosted Enterprise? See Your Own Keys and future Cloud plans · Developer quickstart

Buy → download → install → activate

Three operator steps after a verified Enterprise purchase.

  1. 1
    Purchase Enterprise

    Create a Cloud account with the Enterprise plan and complete verified checkout (identity + payment provider).

  2. 2
    Download your .lic

    After fulfillment, open Licenses and download the signed certificate for this account.

  3. 3
    Install and activate

    Run the installer, open gateway admin, upload the certificate, and activate against the issuer. Network reachability required — offline / air-gap activation is not supported.

install.sh
curl -fsSL https://zenithgw.com/install.sh | bash

Flags: bash -s -- --yes · Swarm: --mode swarm

master.key must survive container recreation; persist /var/lib/zenith as a volume.

Enterprise FAQ

Zenith licensing, residency, and activation.

What is a ZenithCoin?
ZenithCoin is our usage token. Different models consume coins at different rates. For example, Qwen might use 0.5x while a frontier model like Opus uses 3.0x. Your 50 ZenithCoins on Pro go further on efficient models.
How do I estimate my monthly cost?
Use the calculator on our landing page. Enter your expected tokens per month and pick a model. The calculator shows exact ZenithCoin burn based on each model's multiplier.
What is the self-hosted Enterprise option?
Enterprise is an annual self-hosted Zenith Gateway license. Assisted Enterprise includes onboarding and support response targets; Enterprise Early is self-serve activation. Current prices are shown on /pricing in your billing currency. You get a signed .lic for the term, complete control of inference on your network, and no usage metering on the customer-hosted data plane. Hard air-gap and published uptime SLAs are not included.
How do I install self-hosted Zenith?
Run: curl -fsSL https://zenithgw.com/install.sh | bash. The installer pulls stack files from zenithgw.com, starts Docker Compose (or --mode swarm), and uses the Zenith gateway image from ghcr.io/vestavision/zenith-gateway (override with ZENITH_GATEWAY_IMAGE for a private/Enterprise registry). After purchase, download your .lic from Cloud, upload it on the gateway admin activate page, and activate against the issuer. Network reachability is required — fully air-gapped offline activation is not supported. Manual Compose/Swarm, env, ports, and volumes: docs/self-hosted-install.md.
What is post-quantum encryption?
Zenith signs licenses, activation tickets, and managed policy snapshots with NIST ML-DSA-65, and wraps selected secrets at rest with NIST ML-KEM-768 + AES-256-GCM. Public HTTPS should terminate on an edge proxy that negotiates a hybrid PQ KEM (for example X25519MLKEM768). Gateway→issuer activation and policy sync require HTTPS in production. See /docs via the operator handbook post-quantum checklist.
Is my API data logged or stored?
On Zenith Cloud, request metadata is logged for billing and health monitoring, but request/response bodies are never persisted. On self-hosted Enterprise, prompts and responses are not sent to Zenith Cloud. The gateway still needs issuer reachability for license activation and (when enrolled) managed policy sync; set ZENITH_TELEMETRY=off to disable NATS publish and Slack/PagerDuty alert egress. Upstream provider BaseURLs are operator-chosen.
Can I upgrade from Pro to Max?
Yes. Starting a paid plan change keeps your current plan active until the new checkout is verified, then charges the full monthly price of the new plan and applies that plan's monthly ZenithCoin grant. Unused time on the previous plan is not prorated or refunded.
What happens if I run out of ZenithCoins?
Your API requests will return a 402 Payment Required status. From Billing you can buy a one-time ZenithCoin top-up without changing your plan, or wait for the next monthly grant on a paid subscription.
How does Zenith handle provider failures?
Our smart router monitors every provider in real-time. If a provider is slow or returns errors, traffic is instantly rerouted to the next healthy candidate in the pool. You get the fastest successful response without manual intervention.
What is the difference between Auto and explicit model selection?
Auto lets Zenith pick the best provider/model for your request based on real-time health, cost, and latency. Explicit selection gives you full control. Auto comes with a built-in discount because it helps us balance load.
Which APIs does Zenith Cloud expose?
Hosted Cloud focuses on model listing plus chat completions and responses under /api/v1. The self-hosted Gateway exposes the broader OpenAI-compatible /v1 surface (embeddings, images, audio, and more depending on configured providers). Do not assume Cloud and Gateway offer identical endpoints.
What is the Your Own Keys plan?
Your Own Keys is a paid monthly Cloud plan. Current price is shown on /pricing in your billing currency. You add your OpenAI, Anthropic, or Gemini API keys under Provider Keys. Zenith encrypts them at rest and routes chat through Cloud using your credentials. Provider usage bills your provider account; ZenithCoins are not charged for BYOK inference. This is not self-hosted Gateway residency — traffic still enters Zenith Cloud, and request bodies are not persisted.

Ready to run Zenith Gateway?

Purchase Enterprise, download your certificate, install, and activate. Inference stays on your gateway path.