Product — CIPHER UX
The cockpit for your AI traffic.
One console for model access, routing policy, spend analytics, and team governance — sitting in front of a single OpenAI-compatible endpoint.
1
Endpoint
OpenAI-compatible
49
Models
behind the same key
15
Providers
one routing policy
0
Code changes
to adopt
The tour
Policy
You decide which model may answer.
Routing is a policy file, reviewed like code. Cost ceilings, residency rules, and fallbacks live in one place instead of being scattered across services.
Routing policy
4 rules · 1 endpoint
anthropic/claude-opus-5.5
62%
deepseek/deepseek-v4-flash
24%
meta-llama/llama-4-70b
9%
qwen/qwen3-235b
5%
Spend
Cost is visible while it is being spent.
Every request is attributed to a team and a key. The number on this screen is the number that arrives on the invoice — no month-end reconciliation.
Spend by day
last 7 days · EUR
Trace
Every request leaves a readable trail.
Model, status, latency, tokens. When someone asks what was sent and to whom, the answer is one filter away rather than a support ticket.
Request log
live tail
Workflow
From key to production in an afternoon
01
Connect
Point the SDK at your gateway URL and paste the key. No application rewrite.
02
Set policy
Decide which models may serve which data — cost, latency, or residency first.
03
Ship
Turn the feature on. Routing, caching, and failover happen behind the same endpoint.
04
Review
Read the trace: what was asked, which model answered, what it cost.
Inside
Everything in one place
Dashboard
Live traffic, spend, and latency across every model and team — one screen, no export.
Playground
Prompt any model, compare outputs side by side, and promote the winner to production.
Model manager
Enable, version, and route models with policy rules instead of code changes.
Keys and vault
Scoped keys, per-key quotas, and encrypted storage for upstream credentials.
Usage and billing
Per-token cost attribution by team, project, and key.
Logs and traces
Every request filterable by model, status, latency, and token count.
Policies
Routing rules in YAML, versioned in git and reviewed like code.
Access control
SSO, roles, and per-project budgets on the Enterprise plan.
Before / after
What the gateway removes
| Without a gateway | With CIPHER UX | |
|---|---|---|
| Keys in production | One per vendor, scattered in env vars | One scoped key, vaulted upstreams |
| Spend visibility | Reconstruct from invoices | Per key, per team, per day |
| Provider outage | Feature goes down | Automatic fallback |
| Prompt data policy | Enforced by convention | Enforced by routing rules |
| Model change | Code change and redeploy | Policy change |
| Cost per feature | Unknown | Attributed |
Comparison describes the intended operating model, not measured savings. Your numbers depend on traffic shape.
Plans
Start free, scale when you do
Community
Free
One gateway node, core routing, community support.
Pro
€249/mo
High availability, semantic cache, PII guardrails, analytics, email support.
Enterprise
Custom
Air-gapped deployment, SSO/SAML, 99.9% SLA, named engineer.
Feature matrix
| Community | Pro | Enterprise | |
|---|---|---|---|
| Unified endpoint | ✓ | ✓ | ✓ |
| Routing policies | Basic | Advanced | Advanced |
| Semantic cache | — | ✓ | ✓ |
| PII redaction | — | ✓ | ✓ |
| SSO / SAML | — | — | ✓ |
| Air-gapped deploy | — | — | ✓ |
| SLA | — | 99.5% | 99.9% |
| Support | Community | Named engineer |
Console seats are unlimited on every plan. Inference is billed per token at catalog rates — the plan covers the control plane.
Honest note
FAQ
Before you ask
Is CIPHER UX a separate product or part of the API?
Both. The console is the control plane; the gateway is the data plane it configures. You can use the API without ever opening the console, and the console without the gateway only manages keys and spend.
Where does the gateway run?
Inside your network — a VPC, an on-premises cluster, or the edge — as a single container with a Helm chart. State (keys, budgets, logs) stays in your database.
Does routing add latency?
Policy evaluation is in-process and adds single-digit milliseconds. The dominant term remains the upstream model's own time to first token.
Can I keep using my existing provider keys?
Yes. Upstream credentials live in the encrypted vault and never reach client applications. You can mix your own contracts with CIPHER-provided capacity.
What happens if the gateway goes down?
Run it in high-availability mode behind a load balancer, or fall back to a direct CIPHER endpoint. Both paths use the same API schema, so clients do not change.
Take the cockpit for a spin.
Request a demo with your own workload, or launch the console and explore it with sample data.