@agentlens-sdk/sdk

TypeScript/Node.js SDK for AgentLens — real-time observability and debugging for AI agents.

Install

npm install @agentlens-sdk/sdk

With peer dependencies for your AI provider:

npm install @agentlens-sdk/sdk openai                  # for OpenAI
npm install @agentlens-sdk/sdk @anthropic-ai/sdk       # for Anthropic

Prerequisites

Start the AgentLens server first:

pip install agentlens-server
agentlens-server
# Server running at http://localhost:8766

Then open the dashboard at http://localhost:5173.

Usage

Option 1: Auto-instrument (recommended)

import { autoInstrument } from '@agentlens-sdk/sdk';

autoInstrument(); // patches OpenAI + Anthropic Node SDKs

// All subsequent calls are traced automatically
import OpenAI from 'openai';
const client = new OpenAI();
const response = await client.chat.completions.create({ ... });

Option 2: trace() wrapper

import { init, trace } from '@agentlens-sdk/sdk';

init({ serverUrl: 'ws://localhost:8766/ws' });

const myAgent = trace(
  async (query: string) => {
    const result = await callLLM(query);
    return result;
  },
  { name: 'my_agent' }
);

Configuration

import { init } from '@agentlens-sdk/sdk';

init({
  serverUrl: 'ws://localhost:8766/ws',   // AgentLens server WebSocket URL
  httpUrl: 'http://localhost:8766',       // AgentLens server HTTP URL
  sessionName: 'my-agent-v2',            // Optional: name shown in dashboard
});

Design Principles

• Never throws — all SDK operations are wrapped in try/catch. Your agent always runs.
• Non-blocking — async, buffered batch sends. Minimal overhead.
• Graceful degradation — if the server isn't running, events are silently dropped.

License

MIT