@brightdata/ai-sdk

Give your AI agent eyes on the web.

Plug-and-play Vercel AI SDK tools powered by Bright Data - scrape any site, search any engine, and collect structured data from Amazon, LinkedIn, Instagram, Facebook, and more.

Why?

LLMs are powerful reasoners, but they can't see the live web. This package bridges that gap — each export is a ready-made tool() that any AI SDK-compatible model can call autonomously:

• Scrape — fetch any webpage as clean markdown, bypassing anti-bot protection
• Search — query Google, Bing, or Yandex and get structured results
• Datasets — pull product details, job listings, social profiles, and more

No browser automation. No proxy configuration. Just import, configure, and let the model decide when to use each tool.

Quick Start

1. Install

npm install @brightdata/ai-sdk ai zod

2. Get your API key

Sign up at brightdata.com and grab your API key from the dashboard.

Set it as an environment variable, or place it in .env at your project root:

BRIGHTDATA_API_KEY="your_api_key"

If you prefer shell exports:

export BRIGHTDATA_API_KEY="your_api_key"

3. Run your first agent

import { anthropic } from '@ai-sdk/anthropic';
import { generateText, stepCountIs } from 'ai';
import { scrape, search } from '@brightdata/ai-sdk';

const { text } = await generateText({
    model: anthropic('claude-sonnet-4-5-20250929'),
    tools: {
        scrape: scrape(),
        search: search(),
    },
    stopWhen: stepCountIs(5),
    prompt: 'Find and summarize the top 3 trending AI papers this week.',
});

console.log(text);

The model will autonomously decide when to search the web and when to scrape specific pages.

Available Tools

| Tool | Description | Key Inputs | |---|---|---| | scrape | Scrape any URL to clean markdown or HTML | url, country | | search | Search Google, Bing, or Yandex | query, search_engine, country | | amazon_product | Get Amazon product details, pricing, and reviews | url, zipcode | | linkedin_profile | Collect LinkedIn profile data | urls (array) | | linkedin_jobs | Discover LinkedIn job postings | location, keyword, country | | instagram_profile | Fetch Instagram profile info and posts | url | | facebook_profile | Collect Facebook profile data | url | | chatgpt | Collect ChatGPT responses with optional web search | prompt, web_search |

Usage Examples

Web Research Agent

Combine scrape and search to build an agent that can research any topic:

import { openai } from '@ai-sdk/openai';
import { generateText, stepCountIs } from 'ai';
import { scrape, search, amazon_product } from '@brightdata/ai-sdk';

const { text } = await generateText({
    model: openai('gpt-4o'),
    tools: {
        scrape: scrape(),
        search: search(),
        amazon_product: amazon_product(),
    },
    stopWhen: stepCountIs(10),
    prompt: 'Search for the best noise-cancelling headphones and compare'
        + ' prices on Amazon',
});

Streaming with Next.js App Router

// app/api/research/route.ts
import { anthropic } from '@ai-sdk/anthropic';
import { streamText, stepCountIs } from 'ai';
import { scrape, search } from '@brightdata/ai-sdk';

export async function POST(req: Request) {
    const { prompt } = await req.json();

    const result = streamText({
        model: anthropic('claude-sonnet-4-5-20250929'),
        tools: {
            scrape: scrape(),
            search: search(),
        },
        stopWhen: stepCountIs(5),
        prompt,
    });

    return result.toDataStreamResponse();
}

LinkedIn Job Discovery

import { generateText, stepCountIs } from 'ai';
import { anthropic } from '@ai-sdk/anthropic';
import { linkedin_jobs, linkedin_profile } from '@brightdata/ai-sdk';

const { text } = await generateText({
    model: anthropic('claude-sonnet-4-5-20250929'),
    tools: {
        linkedin_jobs: linkedin_jobs(),
        linkedin_profile: linkedin_profile(),
    },
    stopWhen: stepCountIs(5),
    prompt: 'Find senior software engineer jobs in San Francisco'
        + ' and show me the profiles of people in similar roles.',
});

Social Media Intelligence

import { generateText, stepCountIs } from 'ai';
import { anthropic } from '@ai-sdk/anthropic';
import { instagram_profile, facebook_profile } from '@brightdata/ai-sdk';

const { text } = await generateText({
    model: anthropic('claude-sonnet-4-5-20250929'),
    tools: {
        instagram: instagram_profile(),
        facebook: facebook_profile(),
    },
    stopWhen: stepCountIs(5),
    prompt: 'Analyze the social media presence for @natgeo on Instagram.',
});

Cherry-Picking Individual Tools

You don't have to use all tools at once — import only what you need:

import { scrape } from '@brightdata/ai-sdk';

const scrape_tool = scrape({
    api_key: 'your_api_key',
    data_format: 'html',
    country: 'us',
});

Configuration

Every tool factory accepts an options object. Common options:

| Option | Type | Description | |---|---|---| | api_key | string | Bright Data API key. Falls back to BRIGHTDATA_API_KEY from process env or .env file. | | data_format | string | Output format ('markdown', 'html'). Varies by tool. | | country | string | Two-letter country code for geo-targeted requests. |

Tool-Specific Options

search() also accepts:

• search_engine — 'google' | 'bing' | 'yandex' (default: 'google')

linkedin_profile() and facebook_profile() also accept:

• format — 'json' | 'jsonl'

chatgpt() also accepts:

• format — 'json' | 'jsonl'

Requirements

• Node.js 18+
• Vercel AI SDK v5 or v6
• Zod v4+
• A Bright Data API key

Related

• @brightdata/sdk — The core Bright Data SDK this package is built on
• Vercel AI SDK — The AI framework these tools plug into
• Bright Data Docs — Full integration guide

License

MIT - Bright Data