Index is the open-source state-of-the-art AI browser agent that can autonomously perform numerous tasks with a single prompt in the browser environment.

Index is available as a Python package or as an API on the Laminar platform.

​
Use Cases

Index browser agent is ideal for:

  • Web research and data collection
  • Automated testing of web applications
  • Content monitoring and summarization
  • Competitive analysis and market research
  • Customer support automation
  • E-commerce operations

By combining the power of large language models with browser automation capabilities, Index enables you to perform sophisticated web-based workflows that previously required complex custom code with a single API call.

​
Index API

​
Getting Started

Installation

First, install the Laminar TypeScript SDK:

npm install @lmnr-ai/lmnr

Initialize Laminar with your project API key and create a client instance. You can also set the LMNR_PROJECT_API_KEY environment variable instead of passing it directly.

import { Laminar as L, LaminarClient } from '@lmnr-ai/lmnr';

// Initialize Laminar with your project API key
L.initialize({ projectApiKey: "YOUR_PROJECT_API_KEY" });

// Create a client instance
const client = new LaminarClient({
  projectApiKey: "YOUR_PROJECT_API_KEY",
});

​
Using Index Browser Agent

Basic Usage

To run a browser agent that performs tasks based on your prompt:

import { LaminarClient } from '@lmnr-ai/lmnr';

const client = new LaminarClient({
  projectApiKey: "YOUR_PROJECT_API_KEY",
});

// Execute a browser agent task
async function runBrowserAgent() {
  const response = await client.agent.run({
    prompt: "Search for the latest news about AI, find the most recent article, and summarize it.",
  });

  // Print the final result
  console.log(response.result.content);
}

runBrowserAgent();

Streaming Mode

To receive real-time updates as the browser agent executes:

import { LaminarClient } from '@lmnr-ai/lmnr';

const client = new LaminarClient({
  projectApiKey: "YOUR_PROJECT_API_KEY",
});

async function runStreamingBrowserAgent() {
  const response = await client.agent.run({
    prompt: "Lookup the current weather for New York, London, and Tokyo and compare them.",
  });
  
  for await (const chunk of response) {
    if (chunk.chunkType === 'step') {
      // Print step-by-step actions as they happen
      console.log(`Step: ${chunk.summary}`);
    } else if (chunk.chunkType === 'finalOutput') {
      // Print the final result
      console.log(`Final output: ${chunk.content.result.content}`);
    }
  }
}

runStreamingBrowserAgent();

​
Tracing with Laminar

One of the most powerful features of using Index through Laminar is the comprehensive tracing and observability it provides. When you run browser agent tasks through Laminar, you get:

  1. Full Session Replay: Watch exactly what the browser agent saw and did during execution
  2. Synchronized Traces: See the agent’s internal reasoning process alongside browser actions
  3. Performance Metrics: Track duration, token usage, and other performance indicators
  4. Error Analysis: Easily identify and debug failures or unexpected behaviors

To ensure your browser agent sessions are properly traced, make sure you’ve initialized Laminar at the start of your application:

import { Laminar as L } from '@lmnr-ai/lmnr';

L.initialize({ projectApiKey: "YOUR_PROJECT_API_KEY" });

All browser agent activities will automatically be traced and recorded in the Laminar platform, where you can access the session replays and analysis.

​
Visualizing Browser Agent Sessions

The Laminar platform provides a comprehensive visualization interface where you can:

  • Watch time-synced recordings of what the browser agent saw and interacted with
  • Follow the agent’s thought process and decision-making
  • Review extracted data and final outcomes
  • Debug issues by identifying exactly where and why things went wrong
  • Speed up session replays up to 8x for faster debugging

This visual replay gives you unprecedented insight into how the browser agent interprets and executes instructions, making it easier to refine your prompts and build robust web automation workflows.