No proxying
Prompts and completions never touch our servers. Zero added latency.
Use the AI power you already pay for. One plug, every app.
npm install @useoutlet/sdk
Early access for apps and users. One email when the vault opens. No spam.
Your users already pay for AI, whether that's a ChatGPT plan, a Claude account, or an API balance somewhere. Outlet plugs that account into your app as a scoped, spend-capped, instantly revocable key.
Credentials never touch your servers. We never enter the data path. Your app calls the provider directly, on the user's billing.
Works with AI labs
Plugged in straight from your users’ own accounts.
Add the SDK, register your app, get an app_id.
# free while in development
npm install @useoutlet/sdk
Your user clicks once, approves a monthly cap, and Outlet provisions a key scoped to your app only, inside the user's own provider account.
import Outlet from "@useoutlet/sdk"; const session = await Outlet.connect({ appId: "app_yourapp", providers: ["openai", "anthropic"], });
Nothing else changes. Outlet meters spend in the background and kills the key at the cap, so your app and your users are both protected.
const ai = new OpenAI({ apiKey: session.keys.openai });
// direct to the provider, billed to your user, capped by Outlet
Prompts and completions never touch our servers. Zero added latency.
We never resell usage. Users pay their provider, exactly as before.
Apps get per-app capped keys. Users revoke any app in one click.
Free for end users, forever. Free for developers while we build.
The protocol spec is public and the SDK is open source (MIT). The SDK works today in direct mode: your user brings their own key, validated locally (admin keys refused), same one-line API. When the hosted vault opens, the same code upgrades to scoped, spend-capped, revocable keys. The OpenAI adapter works end to end, Anthropic next. Want early access for your app, or a spot in the launch directory of Outlet-enabled apps? Say hello.