@aethermind/agent

Lightweight SDK for real-time AI API cost monitoring. Works with any setup -- official SDKs, raw fetch(), or any HTTP client that calls OpenAI or Anthropic APIs.

Installation

npm install @aethermind/agent
# or
pnpm add @aethermind/agent
# or
yarn add @aethermind/agent

No peer dependencies required. If you use the official OpenAI or Anthropic Node.js SDKs, the agent will patch those too for deeper telemetry -- but it is entirely optional.

Quick Start

import { initAethermind } from "@aethermind/agent";

// Initialize once at app startup
initAethermind({
  apiKey: process.env.AETHERMIND_API_KEY!,
});

That's it. Every call to api.openai.com or api.anthropic.com is now tracked automatically -- regardless of how you make the request.

Using raw fetch()

// No SDK needed -- fetch() calls are intercepted automatically
const response = await fetch("https://api.openai.com/v1/chat/completions", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${process.env.OPENAI_API_KEY}`,
  },
  body: JSON.stringify({
    model: "gpt-4o",
    messages: [{ role: "user", content: "Hello!" }],
  }),
});

// Costs are tracked in the background
const data = await response.json();
console.log(data.choices[0].message.content);

Using the OpenAI SDK

import OpenAI from "openai";

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

// Automatically tracked via SDK patching
const completion = await openai.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello!" }],
});

Using the Anthropic SDK

import Anthropic from "@anthropic-ai/sdk";

const anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY });

// Automatically tracked via SDK patching
const message = await anthropic.messages.create({
  model: "claude-3-5-sonnet-20241022",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello, Claude!" }],
});

How It Works

The SDK uses three layers to capture telemetry with zero code changes:

1. Fetch interceptor -- Patches globalThis.fetch to detect calls to OpenAI/Anthropic API hostnames. Extracts model, tokens, cost, and latency from the response. Works with any HTTP client that uses fetch() under the hood.

2. SDK patching -- Monkey-patches OpenAI.chat.completions.create and Anthropic.messages.create prototypes for SDK-specific telemetry (covers streaming and other SDK features). Only activates if the SDKs are installed.

3. Batch transport -- Captured events are queued locally and flushed to the Aethermind API in batches (default: every 5 seconds or 50 events). Failed events are stored in a dead-letter queue for retry.

Non-LLM fetch calls (e.g. your own API routes) pass through untouched with zero overhead. All telemetry extraction is wrapped in try/catch -- the SDK never throws errors into your application.

Configuration

import { initAethermind } from "@aethermind/agent";

initAethermind({
  apiKey: "your-aethermind-api-key", // Required
  endpoint: "https://aethermind-agentos-production.up.railway.app", // Optional
  flushInterval: 5000, // Optional: ms between flushes (default: 5000)
  batchSize: 50, // Optional: events per batch (default: 50)
  enabled: true, // Optional: enable/disable (default: true)
});

| Option | Type | Default | Description | | --------------- | --------- | --------------------------- | ---------------------------------- | | apiKey | string | required | Your Aethermind API key | | endpoint | string | https://aethermind-agentos-production.up.railway.app | API endpoint URL | | flushInterval | number | 5000 | Milliseconds between batch flushes | | batchSize | number | 50 | Maximum events per batch | | enabled | boolean | true | Enable/disable monitoring |

Disable Monitoring in Development

initAethermind({
  apiKey: process.env.AETHERMIND_API_KEY!,
  enabled: process.env.NODE_ENV === "production",
});

CommonJS Usage

const { initAethermind } = require("@aethermind/agent");

initAethermind({
  apiKey: process.env.AETHERMIND_API_KEY,
});

Troubleshooting

Events not appearing in dashboard?

1. Wait 5 seconds (default batch interval)
2. Verify API key is correct: process.env.AETHERMIND_API_KEY
3. Check endpoint URL is correct
4. Ensure enabled: true (default)

TypeScript errors?

Make sure you have @types/node installed:

npm install -D @types/node

Want to verify initialization?

The SDK logs to console when initialized:

[Aethermind] SDK initialized successfully

If disabled:

[Aethermind] SDK initialized but disabled

Documentation

• Full Documentation
• API Reference
• Dashboard

Support

• Report Issues
• Discussions
• Email: [email protected]

License

MIT - Aethermind Team