The TokenFence Node.js SDK is now live on npm. One install, zero dependencies, full TypeScript support.

Why Node.js?

When we launched the Python SDK last week, the #1 request was: "When's the Node.js version coming?"

Fair question. A huge chunk of production AI agent code runs on Node.js and TypeScript — Next.js API routes, Express backends, serverless functions, Vercel Edge, Cloudflare Workers. If you're building AI-powered products, chances are your backend is JavaScript.

Today, that's covered.

npm install tokenfence

What You Get

Per-workflow budget caps — set a dollar limit per agent run, per user, per request
Auto model downgrade — when budget gets tight, automatically switch from GPT-4o to GPT-4o-mini
Hard kill switch — exceed the budget? The agent stops. No surprises on your bill.
Full TypeScript types — first-class DX with complete type definitions
ESM + CommonJS — works everywhere: import or require
Zero dependencies — nothing to audit, nothing to break

Quick Start

import { TokenFence } from 'tokenfence';

const fence = new TokenFence({
  budget: 0.50,           // $0.50 per workflow
  downgradeAt: 0.80,      // switch models at 80% budget
  downgradeModel: 'gpt-4o-mini',
  onKill: (usage) => console.log('Budget exceeded:', usage)
});

// Wrap your OpenAI call
const result = await fence.guard(async () => {
  return await openai.chat.completions.create({
    model: fence.currentModel('gpt-4o'),
    messages: [{ role: 'user', content: 'Analyze this data...' }]
  });
});

Works With Everything

TokenFence is provider-agnostic. It works with:

OpenAI — GPT-4o, GPT-4o-mini, GPT-5, o1, o3
Anthropic — Claude 4, Claude Sonnet, Claude Haiku
Google Gemini — via OpenAI-compatible API
Any LLM — if it has a cost, TokenFence can cap it

Python + Node.js = Full Stack Coverage

With both SDKs live, you can protect your entire AI stack:

# Python (FastAPI, Django, LangChain, CrewAI)
pip install tokenfence

# Node.js (Next.js, Express, serverless)
npm install tokenfence

Same concepts, same API design, both languages. Your team uses Python for ML pipelines and Node.js for the web layer? Both are covered.

What's Next

Usage dashboard — see your agent spend in real-time
Team budgets — aggregate limits across multiple workflows
Webhook alerts — get notified before budgets hit
Framework plugins — native LangChain.js and Vercel AI SDK integrations

Get started now:

npm install tokenfence

Read the documentation or check out examples on GitHub.

TokenFence Node.js SDK Is Live on npm