How to Integrate Yutori n1 with Bright Data Browser API

Integrating Bright Data with Yutori gives you an autonomous web research agent powered by the Yutori n1 reasoning model and Bright Data’s Browser API - combining real Chromium browsing with residential IPs for undetectable, scalable web access. The n1-brightdata Python package is the official CLI and library for this integration, including support for:

Autonomous web research - Give the agent a task in plain language. It browses the web, interacts with pages, and returns a synthesized answer.
Bright Data Browser API - A real Chromium browser routed through Bright Data’s residential proxy network, bypassing bot detection, CAPTCHAs, and geo-restrictions automatically.
Yutori n1 model - An advanced reasoning model that decides what browser actions to take (click, scroll, type, navigate) at each step until it has enough information to answer.

How to use n1-brightdata

Obtain Your Credentials

You need credentials from two services:Bright Data:

Log in to your Bright Data dashboard.
Create or open a Browser API zone.
Copy the CDP WebSocket URL - it looks like: wss://brd-customer-XXXXXX-zone-scraping_browser:PASSWORD@brd.superproxy.io:9222

Yutori:

Log in to your Yutori dashboard.
Generate an API key from your account settings.

Install the n1-brightdata Package

Install the package from PyPI:

pip install n1-brightdata

Then install the Chromium browser used by Playwright:

playwright install chromium

Configure Your Credentials

Run the built-in interactive setup wizard:

n1-brightdata setup

This walks you through entering your Bright Data CDP URL and Yutori API key, saves them securely to ~/.n1-brightdata/credentials.json, installs Playwright, and optionally tests your connections.

Or configure manually by creating ~/.n1-brightdata/credentials.json:

{
  "YUTORI_API_KEY": "your_yutori_api_key_here",
  "BRD_CDP_URL": "wss://brd-customer-XXXXXX-zone-scraping_browser:PASSWORD@brd.superproxy.io:9222"
}

You can also pass credentials directly as CLI flags or environment variables:

n1-brightdata "your task" \
  --yutori-api-key "your-key" \
  --brd-cdp-url "wss://..."

Run your research task

The n1-brightdata integration supports three usage patterns:

Basic CLI
Advanced CLI Options
Python API

Reference: Full command options

Run any web research task from your terminal in a single command.

# Ask any question — the agent browses the web to answer it
n1-brightdata "What is the current price of Bitcoin?"

Reference: Tuning parameters

Fine-tune screenshot quality, step limits, and credential sources for your use case.

# Use PNG for visual or layout-inspection tasks
n1-brightdata "Describe the layout of the Airbnb homepage" \
  --url "https://www.airbnb.com" \
  --screenshot-format png

Reference: Programmatic usage

Import and call the agent directly from your Python code.

from n1_brightdata import build_agent_config, run_agent

# Load credentials from ~/.n1-brightdata/credentials.json (or env vars)
config = build_agent_config()

# Run a task
run_agent(
    task="What are the key differences between GPT-4o and Claude?",
    start_url="https://www.google.com",
    config=config,
    max_steps=30,
)

Command Reference

n1-brightdata [TASK] [OPTIONS]

Option	Type	Default	Description
`TASK`	string	(required)	The research task for the agent to complete
`--url`	string	`https://www.google.com`	Starting URL before the agent loop begins
`--max-steps`	integer	`30`	Maximum number of browser actions (min: 1)
`--screenshot-format`	`jpeg` \| `png`	`jpeg`	Format of screenshots sent to the model
`--jpeg-quality`	integer	`60`	JPEG quality when format is `jpeg` (1–100)
`--screenshot-timeout-ms`	integer	`90000`	Screenshot timeout in milliseconds
`--yutori-api-key`	string	(env)	Yutori API key (overrides env / credentials file)
`--brd-cdp-url`	string	(env)	Bright Data CDP WebSocket URL (overrides env / credentials file)
`--env-file`	path	`./.env`	Custom path to a `.env` file

Optional Tuning

Place these in a .env file in your project directory (or export them as shell variables) to tune agent behavior without touching your credentials:

N1_SCREENSHOT_FORMAT=jpeg          # jpeg | png
N1_JPEG_QUALITY=60                 # 1–100
N1_SCREENSHOT_TIMEOUT_MS=90000     # screenshot timeout in milliseconds
N1_MAX_REQUEST_BYTES=9500000       # trim old screenshots if payload exceeds this
N1_KEEP_RECENT_SCREENSHOTS=6       # how many screenshots to keep in context
N1_ENABLE_SUFFICIENCY_CHECK=true   # stop early when the agent is confident
N1_STOP_CONFIDENCE_THRESHOLD=0.78  # confidence threshold for early stopping (0–1)

Configuration Precedence

CLI flags  >  Shell env vars  >  .env file  >  ~/.n1-brightdata/credentials.json

Programmatic API

The package also exposes a Python API for embedding the agent in your own code:

from n1_brightdata import AgentConfig, build_agent_config, run_agent

Symbol	Description
`AgentConfig`	Dataclass holding all configuration (credentials + tuning)
`build_agent_config()`	Load config from credentials file, env vars, and `.env`
`run_agent(task, start_url, config, max_steps)`	Run the agentic loop and print the result

Introduction

Quick Start

For AI Agents

Use Cases

Integrations

How to Integrate Yutori n1 with Bright Data Browser API

How to use n1-brightdata

Command Reference

Optional Tuning

Configuration Precedence

Programmatic API

Introduction

Quick Start

For AI Agents

Use Cases

Integrations

​How to use n1-brightdata

​Command Reference

​Optional Tuning

​Configuration Precedence

​Programmatic API

How to use n1-brightdata

Command Reference

Optional Tuning

Configuration Precedence

Programmatic API