Model Catalog API
Embed Cencori's public model and provider catalog — with logos — inside your own product. No API key required.
Overview
The catalog API ships the same data rendered on cencori.com/ai-gateway/models: every provider, every model, context windows, capabilities, and display pricing — plus hotlinkable provider logos.
All three endpoints are public and keyless. Call them from a browser bundle safely; never put a secret API key in client-side code to fetch the catalog.
Responses are cached (s-maxage=3600), so refetch at most hourly.
List models
curl "https://cencori.com/api/models?provider=openai&type=chat"Query params (all optional, combinable):
Each model row looks like this:
{
"id": "gpt-5",
"name": "GPT-5",
"provider": "openai",
"provider_name": "OpenAI",
"provider_logo_url": "https://cencori.com/api/providers/openai/logo",
"type": "chat",
"types": ["chat"],
"context_window": 400000,
"context_display": "400K",
"description": "Flagship model",
"capabilities": { "tools": true },
"pricing": {
"input_per_million": 5,
"output_per_million": 15,
"currency": "USD",
"input_display": "$5.00",
"output_display": "$15.00",
"per_image": false,
"per_image_usd": null
}
}Image models bill per image instead: per_image: true with per_image_usd (e.g. 0.02) and "output_display": "per img".
List providers
curl "https://cencori.com/api/providers"{
"object": "list",
"data": [
{
"id": "openai",
"name": "OpenAI",
"logo_url": "https://cencori.com/api/providers/openai/logo",
"website": "https://openai.com",
"docs_url": "https://platform.openai.com/docs",
"model_count": 32
}
],
"total": 13
}Logos
Use logo_url (or per-model provider_logo_url) directly as an image source — it always returns an SVG for known providers and 404s only for unknown ids:
<img src="https://cencori.com/api/providers/openai/logo" width="20" height="20" alt="OpenAI" />Two things to know:
- Vendors with a checked-in brand mark serve the real logo; the rest serve a neutral monogram tile so your
<img>never breaks. - Several brand marks are white (OpenAI, Anthropic). Render them on a dark chip or they disappear on light backgrounds.
SDK usage
Install cencori@^1.8.0. The catalog helpers are keyless — no client or API key needed, so they are browser-safe:
import { listCatalogModels, listCatalogProviders, providerLogoUrl } from 'cencori';
// or: import { ... } from 'cencori/catalog';
const { data: models } = await listCatalogModels({ provider: 'openai' });
const { data: providers } = await listCatalogProviders();
const logo = providerLogoUrl('openai'); // string, no fetchReact example — a provider shelf with logos:
import { useEffect, useState } from 'react';
import { listCatalogProviders, type CatalogProvider } from 'cencori';
export function ProviderShelf() {
const [providers, setProviders] = useState<CatalogProvider[]>([]);
useEffect(() => {
listCatalogProviders().then((res) => setProviders(res.data));
}, []);
return (
<ul>
{providers.map((p) => (
<li key={p.id}>
<img src={p.logo_url} width={20} height={20} alt={p.name} />
{p.name} · {p.model_count} models
</li>
))}
</ul>
);
}Public catalog vs authenticated models
Use the public surface for browsing and display. Use /v1/models server-side when you need to know what a specific key can actually call.