// how it works

Up and saving in 5 minutes.

Three steps. No infrastructure changes. No model retraining.

01

Drop in the SDK

Install the PromptForge package and wrap your existing LLM calls. Two lines of code. No rewrite required.

step-01.sh
npm install @promptforge/sdk

// Before
const res = await openai.chat.completions.create({ ... });

// After — same call, now routed & compressed
import { pf } from "@promptforge/sdk";
const res = await pf.chat({ ... });
02

Connect your providers

Paste your OpenAI, Anthropic, and Google API keys in the dashboard. PromptForge proxies requests and keeps your keys encrypted at rest.

step-02.sh
# .env
PROMPTFORGE_API_KEY=pf_live_...

# Supports
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_AI_API_KEY=AIza...
03

Watch savings roll in

The Cost Dashboard updates in real time. See exactly which tool saved what, down to the feature and provider. No spreadsheets.

step-03.sh
// Cost Dashboard — live
{
  "month_to_date": {
    "spend": 1842.31,
    "saved": 4231.00,
    "reduction_pct": 52.4
  },
  "top_saver": "prompt_compressor",
  "tokens_saved": 3_210_000
}

Built for production from day one.

⚡

~12ms overhead

PromptForge adds negligible latency. The compression alone more than covers it.

🔒

Keys never leave your account

We encrypt provider keys at rest, isolate them per-account, and never log request payloads.

🔌

Works with any framework

Node.js, Python, Go, or plain HTTP. If you can call an API, you can use PromptForge.

📊

Realtime attribution

Every dollar of savings is attributed to the specific tool, feature, and provider that produced it.

Ready to drop in?

Free plan. No credit card. Savings start the moment you connect your first provider.