Question 1

How does Cencori actually work?

Accepted Answer

Cencori sits between your app and every AI provider. Every request goes through Cencori first — where it's checked for security threats, routed to the right model, logged for compliance, and tracked for cost. The response comes back through the same layer, filtered and audited. You get full visibility and control without changing how you call the AI.

Question 2

What AI models and providers do you support?

Accepted Answer

We support 14+ AI providers with 75+ models including: OpenAI (GPT-5, GPT-4o, o3), Anthropic (Claude Opus 4, Sonnet 4), Google (Gemini 3 Pro, Gemini 2.5 Flash), xAI (Grok 4), Mistral (Large 3), DeepSeek (V3.2), Cohere (Command A), Perplexity (Sonar Pro), and more. You can switch between models dynamically - no code changes needed.

Question 3

What's the latency overhead?

Accepted Answer

Typically 10-50ms added latency for security checks. Our servers are globally distributed on Vercel's edge network, so requests are routed from the nearest location. For most use cases, this is imperceptible - the AI model generation time (often 1-5 seconds) dominates total latency.

Question 4

What's the difference between Production and Development environments?

Accepted Answer

Each project has separate Production and Development environments with isolated API keys. Production keys (start with cen_) count toward your tier limits and are logged for compliance. Development keys (start with cen_test_) don't count toward limits, perfect for testing. This lets you iterate safely without affecting production quotas or compliance logs.

Question 5

What happens when I hit my request limit?

Accepted Answer

We return a 429 error with details about your usage and upgrade options. Your app can handle this gracefully by showing users a message or queuing requests. You can upgrade your plan mid-month instantly - higher limits apply immediately, and we prorate the difference.

Question 6

Is my data secure? Where do you store requests and responses?

Accepted Answer

Yes. All data is encrypted in transit (TLS 1.3) and at rest (AES-256). We store request/response logs in Supabase (SOC 2 compliant) for audit and analytics purposes. Logs are retained for 90 days by default (customizable for enterprise). We never train models on your data or share it with third parties.

Question 7

Can I use this for just one project or do I need to migrate everything?

Accepted Answer

Cencori is project-scoped, so you can start with just one app or feature. Each project gets its own API keys, quotas, and analytics. No need for an all-or-nothing migration - add projects as you need them. Many customers start with their highest-risk or highest-value feature first.

Question 8

How is Cencori different from using OpenAI directly?

Accepted Answer

OpenAI gives you the model. Cencori gives you control over what happens with it. Security (prompt injection, PII detection, content filtering), multi-provider routing (switch between GPT-4, Claude, Gemini without code changes), audit logs and compliance (SOC 2, GDPR), and cost tracking per user or feature. Think of it as the difference between a raw database and Supabase — same power, but with everything production actually needs built in.

Question 9

What is Cencori?

Accepted Answer

Cencori is the control layer for AI in production. The moment you ship AI to real users, you need to know what data is leaving your system, who's trying to manipulate your AI, how much it's costing per user, and what to do when a provider goes down. Cencori answers all of those questions — automatically, from your first request.

Every AI request, under your control.

Everything you need to ship AI.

AI Gateway

Infrastructure in minutes

Install the SDK

Get your API key

Start building

Universal AI Gateway

See everything. Miss nothing.

Know what you're spending. Everywhere.

Set limits. Sleep better.

Frequently Asked Questions

Your AI is live. Do you know what's happening inside it?