CenovaGate
Open console

OpenAI-compatible AI gateway

Connect once. Optimise every AI request.

Put CenovaGate between your application and AI providers without rewriting your request format. The CenovaGate Decision Engine checks your policies, avoids unnecessary calls, reduces context, and routes each request to the best eligible model—with a quality-preserving fallback when something is unavailable.

Create your free workspace Run a free AI Cost Audit

First request

Four steps to connect

Your existing OpenAI-compatible request can stay the same. Change the connection details, then let CenovaGate handle the decision path.

  1. STEP 1

    Connect a provider

    Choose a supported AI provider and add its credential securely.

  2. STEP 2

    Create an application

    Add your application and choose its Development, Staging, or Production environment.

  3. STEP 3

    Generate a CenovaGate key

    Create a gateway key for that environment. It is shown once, so save it securely.

  4. STEP 4

    Change the API base URL

    Keep your OpenAI request shape and point it to CenovaGate instead.

Send a request

Use a placeholder while testing your integration. Never put a real gateway key in browser code or a public repository.

curl https://cenovagate.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_CENOVAGATE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4.1-mini",
    "messages": [
      { "role": "user", "content": "Summarise this support request." }
    ]
  }'

Every request, in order

A decision path built around quality

  1. 01Incoming request
  2. 02Organisation policies
  3. 03Cache or deterministic answer
  4. 04Context optimisation
  5. 05Best eligible model
  6. 06Quality-preserving fallback

An incoming request is checked against organisation policies, then safe cached or deterministic answers, context optimisation, the best eligible model, and finally a quality-preserving fallback if required.

Examples

The right path for the work

Your organisation’s policies always run first. Automated routing works only with models that meet your availability, capability, quality, context, latency, and budget requirements.

Repeated question

Safely cached answer

Reuse an exact, eligible response instead of paying for another model call.

Simple support classification

Lower-cost eligible model

Automated routing can select a capable lower-cost model within your rules.

Complex reasoning

Premium eligible model

Higher-complexity work stays with a model suited to the quality required.

Provider or optimisation issue

Configured fallback or requested model

The failed layer is bypassed so quality and availability remain the priority.

Built for real-time routing

Fast routing decisions, built for live AI requests.

CenovaGate evaluates the best eligible execution path in milliseconds, then lets the selected AI model do the work.

  • Bounded decision time

    Routing decisions run within a strict time budget, so the gateway never holds a request hostage while it evaluates the best path.

  • Quality-preserving fallback

    If an optimisation layer or provider is unavailable, it is bypassed safely and your request still reaches a model that can answer it well.

  • No unnecessary AI calls

    Cached answers, deterministic responses, and trimmed context mean you pay only for the AI work that is genuinely needed.

Decision timing is reported separately from total AI response time. Once live production monitoring provides measured median and 95th-percentile decision latencies, they will be published here as CenovaGate production measurements.

You set the level of control

Start by observing. Increase automation when ready.

Every new organisation begins in Observe mode, so CenovaGate measures opportunities without changing live traffic.

  1. 1 / 5Default

    Observe

    Measure requests and opportunities without changing traffic.

  2. 2 / 5

    Shadow

    Compare an alternative privately while your chosen model still answers.

  3. 3 / 5

    Recommend

    Show routing recommendations while your team keeps control.

  4. 4 / 5

    Optimise

    Apply approved caching, context, and routing changes within your rules.

  5. 5 / 5

    Autonomous

    Adapt routing automatically within your organisation’s safeguards.

See where your AI spend can work harder.

Create a workspace to connect your first request, or start with an audit using anonymised request metadata.

Create your free workspace Run a free AI Cost Audit