Real products already ship on withConflux — LLM apps on the gateway (their own OpenAI key or managed credits) and a managed SEO directory. However each one is built, it runs without the busywork.
Client names are held back on purpose — but every capability below is real, taken straight from how each one actually runs.
BYOK · own OpenAI key
A developer platform, billed on its own key
A SaaS product with an in-app AI assistant
Wanted spend caps, usage analytics and EU-hosted logging — without moving billing off its own OpenAI account. The gateway sits in front; the key stays theirs.
Kept the official OpenAI SDK — no call sites rewritten
A runtime Direct ↔ Gateway toggle — switch with no redeploy
Billed on their own key — control, not a markup
// the whole change
new OpenAI({
baseURL: "https://withconflux.com/api/v1",apiKey: process.env.CONFLUX_API_KEY,
})
Managed · prepaid credits
A content app with no OpenAI account at all
A social-content generation product
Didn’t want to own an OpenAI relationship — just working AI and one predictable bill. withConflux supplies the key behind the gateway and meters against prepaid credits.
No OpenAI account or key to manage — the gateway holds it
Out of credits returns a clean 402 → graceful fallback
// same two options — key + base URL
new OpenAI({
baseURL: "https://withconflux.com/api/v1",apiKey: process.env.CONFLUX_API_KEY, // managed
})
SEO · managed schema
A trades directory that writes no code at all
A regional home-services directory
Wanted rich structured data on every listing without building or maintaining a schema pipeline. withConflux manages it end to end and pushes it to their own site.
308 listings across 33 service types managed for them
Schema pushed to their own domain — served locally, zero latency
Fully managed — nothing to build or run on their side
Same login adds AI Visibility — track how AI recommends their listings, and fix it
NO CODE
They connect their site once; we keep the structured data current.
Going live
Three steps from your first call.
The same path for the two gateway integrations above — the only fork is which billing mode you pick in step two.
1
Point the SDK at the gateway
Set the base URL and a withConflux key on your existing OpenAI client. Every model call stays exactly as written.
2
Pick a billing mode
Bring your own OpenAI key and keep that bill, or run on managed prepaid credits with auto-reload. Switch later with no code change.
3
Ship with the guardrails on
Spend caps, per-model usage analytics and EU-hosted logging are live from the first request — nothing extra to wire up.
Client identities anonymized on purpose. Every claim on this page is a real capability of the two live integrations — no quotes, ratings or invented metrics.