@dnorio/agent-meter

Lightweight Node.js/TypeScript SDK to track AI agent tool calls, costs, and latency.

Install

npm install @dnorio/agent-meter

From a local tarball:

cd sdk/node && npm ci && npm pack
npm install ./dnorio-agent-meter-0.1.0.tgz

Quick Start (60 seconds)

import { AgentMeter } from "@dnorio/agent-meter";

const am = new AgentMeter({ apiKey: "am_live_..." });

// Track a tool call
const span = am.track("web_search", { model: "gpt-4o" });
const result = await doWebSearch(query);
am.finish(span);

// Track with tokens/cost
const span2 = am.track("code_review", { model: "claude-sonnet-4-20250514" });
span2.estimatedInputTokens = 2000;
span2.estimatedOutputTokens = 500;
span2.usdCost = 0.012;
am.finish(span2);

// Flush happens automatically every 5s
// Force flush:
await am.flush();

Configuration

| Env Variable | Description | Default | |---|---|---| | AGENT_METER_API_KEY | API key (am_live_...) | — | | AGENT_METER_ENDPOINT | Server URL | http://localhost:8081 | | AGENT_METER_IDE | IDE identifier | node-sdk | | AGENT_METER_AGENT | Agent name | — |

Or pass via constructor:

const am = new AgentMeter({
  apiKey: "am_live_...",
  endpoint: "http://localhost:3000",
  ide: "my-custom-agent",
  agent: "research-bot",
});

Conversations & Tasks

const span = am.track("generate_code", {
  model: "claude-sonnet-4-20250514",
  conversationId: "conv-abc123",
  taskId: "implement-feature-x",
});

Protocol

The SDK sends spans as OTLP-compatible JSON to POST /v1/traces.