Quick Start

This guide gets a real Cencori request running. If you already have an app, use Add Cencori to an Existing Product for the dashboard-to-code path.

Cencori sits between your app and model providers, so there are two separate setup pieces:

• Cencori project key: authenticates your app to Cencori. This is the CENCORI_API_KEY value your code uses.
• Provider access: lets Cencori route to OpenAI, Anthropic, Google, or another provider. Use Cencori managed access when it is enabled for your project, or add your own provider key in Project > Providers.

If your first request says Provider 'openai' is not configured, your Cencori key is valid. You still need provider access for that model.

1. Sign up at cencori.com/signup.
2. Create an organization and project.
3. Open the project, then go to API Keys.
4. Generate a secret key for server-side use.

Store the key as an environment variable:

# .env.local for Next.js, .env for most Node apps
CENCORI_API_KEY=csk_...

Keep csk_... keys on the server. Do not put them in browser code or NEXT_PUBLIC_* variables.

Choose the model you want to call, then make sure your project can route to it:

1. Open Models or Playground in your project.
2. Pick one model ID for the first test.
3. If a free, managed, or catalog model is enabled for your project, use that first.
4. If you want BYOK or a custom provider, open Project > Providers and add the provider before testing.

For the fastest first test, use one known model such as gpt-4o, claude-sonnet-4.5, or gemini-2.5-flash.

If you are starting a new app, use the scaffolder:

npx create-cencori-app my-ai-app --template nextjs
cd my-ai-app

Open .env.local, paste your Cencori project key, then run:

npm run dev

If you already have a product, do not start with a blank scaffold. Follow Add Cencori to an Existing Product to wire Cencori into your current backend, OpenAI client, or framework.

The short version:

1. Pick a model in the dashboard.
2. Create a csk_... key in API Keys.
3. Store it as CENCORI_API_KEY on your server.
4. Change your SDK client or base URL.
5. Send one request and confirm it appears in Logs.

This test removes your framework from the equation:

curl https://cencori.com/api/ai/chat \
  -H "CENCORI_API_KEY: $CENCORI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

For OpenAI-compatible clients:

curl https://api.cencori.com/v1/chat/completions \
  -H "Authorization: Bearer $CENCORI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'

All requests are logged in your dashboard with:

• Full prompt and response
• Token usage and costs
• Latency metrics
• Security scan results