AI features on static / JAMstack sites

Add a model-powered feature to a static or JAMstack site while keeping the key server-side, without moving to a full server runtime.

The problem

JAMstack sites are mostly static, so there is no obvious server-side place to keep a secret. Teams reach for serverless functions per call, which adds cold starts and a maintenance surface, or they leak the key into client code.

With Perishable

Deploy the Perishable proxy as a single long-running process alongside the site. Front-end code calls the OpenAI-compatible surface through it; the proxy attaches the real key server-side and rate-limits per fingerprint.

How it flows

Step by step

01 · deploy

Run the proxy as one process behind your normal TLS / reverse proxy.

02 · point

Configure the browser SDK with the proxy URL.

03 · limit

Set per-fingerprint rate limits to bound spend.

04 · iterate

Add model features from the static front end without a per-call function.

What you get

Outcome

AI features on an otherwise static site
No per-request serverless cold starts
Spend bounded by per-fingerprint limits

Try it on your app

Follow the quickstart or read the request flow.