1. Overview & Architecture

2. Quickstart (1-Line Drop-in)

Integrate JevProxy into any existing application by pointing your SDK's baseURL to the JevProxy gateway:

import OpenAI from "openai";

// Drop-in JevProxy: change only baseURL and pass your JevProxy key
const openai = new OpenAI({
  baseURL: "https://api.jevproxy.com/v1",
  apiKey: process.env.JEVPROXY_API_KEY, // e.g. "jev_live_..."
});

// Any decision turns execute in <25ms at $0.0001!
const response = await openai.chat.completions.create({
  model: "gpt-4o",
  messages: [
    { role: "system", content: "Select optimal action: [approve, reject, escalate]" },
    { role: "user", content: "Refund requested on order #891" }
  ]
});

3. Coding Agents & Tool Calling Acceleration

Modern coding agents (Claude Code, Cursor, Codex, Windsurf, OpenCode, Cline) spend huge amounts of time and reasoning tokens on every turn deciding which tool to call. JevProxy sits directly between your agent and upstream LLM providers, intercepting the tools: [...] array and using TypeSafe Jev in sub-25ms to accelerate execution:

4. AI Agent Skill (/skill.md)

Autonomous developer agents (such as Claude Code, Cursor, GitHub Copilot, or LangChain) can natively load JevProxy as an autonomous skill. The specification is hosted publicly at /skill.md and /.well-known/skill.md.

5. System One Decision Primitives

TypeSafe Jev replaces stochastic text generation with calibrated non-autoregressive decision models. Three core primitives are supported:

6. Gateway API Reference

7. Control Headers & Routing Flags

Customize how requests are routed through JevProxy by passing optional HTTP headers:

8. Prompt Injection & Safety Guardrails

Intercept adversarial inputs, jailbreaks, and system prompt override attempts in sub-25ms before any tokens reach an expensive LLM:

// Check incoming prompt safety via JevProxy Guardrails API
const response = await fetch("https://api.jevproxy.com/api/v1/guardrails/check", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer jev_live_your_key_here"
  },
  body: JSON.stringify({
    prompt: userQuery
  })
});

const data = await response.json();
if (data.isThreat) {
  throw new Error("Prompt injection attempt intercepted: " + data.threatLevel);
}

9. Response Headers & Telemetry

Every request proxied through JevProxy includes real-time telemetry headers so you can audit performance on every turn:

10. Full Code Examples

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://api.jevproxy.com/api/v1",
    api_key=os.environ.get("JEVPROXY_API_KEY")
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": "Route to billing, technical, or sales."},
        {"role": "user", "content": "I need to update my company payment method."}
    ]
)

print(response.choices[0].message.content)
# 18.4ms latency | $0.0001 cost

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.jevproxy.com/api/v1",
    api_key=os.environ.get("JEVPROXY_API_KEY")
)

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[{"role": "user", "content": "What is the git status?"}],
    tools=[{"name": "git_status", "description": "Check working tree status", "input_schema": {"type": "object"}}]
)
# Direct execution in <30ms with 0 reasoning tokens burned!

curl https://api.jevproxy.com/api/v1/chat/completions \
  -H "Authorization: Bearer $JEVPROXY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {"role": "system", "content": "Classify priority: p0, p1, or p2."},
      {"role": "user", "content": "Checkout DB replica lag is 40 minutes."}
    ]
  }'