Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/browser-use/browser-use/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

ChatBrowserUse is the recommended LLM provider for Browser Use, offering the fastest and most cost-effective models specifically optimized for browser automation tasks. It achieves 3-5x faster task completion compared to standard models.
Get started with $10 of free LLM credits at cloud.browser-use.com/new-api-key

​
Basic Usage

from browser_use import Agent, ChatBrowserUse
import asyncio

async def main():
    llm = ChatBrowserUse()
    agent = Agent(
        task="Find the number 1 post on Show HN",
        llm=llm,
    )
    await agent.run()

if __name__ == "__main__":
    asyncio.run(main())

​
Configuration

​
Parameters

​
model
str
default:"bu-latest"
Model name to use. Available options:
  • bu-latest or bu-1-0: Default model
  • bu-2-0: Latest premium model
  • browser-use/bu-30b-a3b-preview: Browser Use Open Source Model
​
api_key
str
default:"None"
API key for browser-use cloud. Defaults to BROWSER_USE_API_KEY environment variable.
Get your API key at cloud.browser-use.com/new-api-key
​
base_url
str
default:"None"
Base URL for the API. Defaults to BROWSER_USE_LLM_URL environment variable or https://llm.api.browser-use.com.
​
timeout
float
default:"120.0"
Request timeout in seconds.
​
max_retries
int
default:"5"
Maximum number of retries for transient errors (429, 500, 502, 503, 504).
​
retry_base_delay
float
default:"1.0"
Base delay in seconds for exponential backoff.
​
retry_max_delay
float
default:"60.0"
Maximum delay in seconds between retries.

​
Advanced Usage

​
Custom Model Selection

from browser_use import Agent, ChatBrowserUse

# Use premium model
llm = ChatBrowserUse(model='bu-2-0')

# Use open source model
llm = ChatBrowserUse(model='browser-use/bu-30b-a3b-preview')

agent = Agent(
    task="Complex browser automation task",
    llm=llm,
)

​
With Session ID for Sticky Routing

from browser_use import Agent, ChatBrowserUse

llm = ChatBrowserUse()

agent = Agent(
    task="Task requiring session persistence",
    llm=llm,
)

# The session_id is automatically managed by the agent
# Same session routes to the same container for consistency

​
Custom Base URL

import os
from browser_use import Agent, ChatBrowserUse

# Set via environment variable
os.environ['BROWSER_USE_LLM_URL'] = 'https://custom-endpoint.com'

# Or set directly
llm = ChatBrowserUse(
    base_url='https://custom-endpoint.com',
    api_key='your-api-key'
)

agent = Agent(task="Your task", llm=llm)

​
Environment Setup

.env
BROWSER_USE_API_KEY=your_api_key_here
# Optional: Custom endpoint
BROWSER_USE_LLM_URL=https://llm.api.browser-use.com

​
Error Handling

ChatBrowserUse automatically handles common errors with exponential backoff retry logic:
  • Rate Limits (429): Automatic retry with backoff
  • Server Errors (500, 502, 503, 504): Automatic retry with backoff
  • Network Errors: Automatic retry with backoff
  • Authentication (401): Raised immediately as ModelProviderError
  • Insufficient Credits (402): Raised immediately as ModelProviderError
from browser_use import Agent, ChatBrowserUse
from browser_use.llm.exceptions import ModelProviderError, ModelRateLimitError

try:
    llm = ChatBrowserUse(max_retries=3)
    agent = Agent(task="Your task", llm=llm)
    result = await agent.run()
except ModelRateLimitError as e:
    print(f"Rate limit exceeded: {e.message}")
except ModelProviderError as e:
    print(f"Provider error: {e.message} (Status: {e.status_code})")

​
Properties

​
provider

Returns the provider name: "browser-use" 
llm = ChatBrowserUse()
print(llm.provider)  # "browser-use"

​
name

Returns the model name. 
llm = ChatBrowserUse(model='bu-2-0')
print(llm.name)  # "bu-2-0"

​
Methods

​
ainvoke()

Asynchronously invoke the model with messages. 
from browser_use.llm.messages import SystemMessage, UserMessage

llm = ChatBrowserUse()

messages = [
    SystemMessage(content="You are a helpful assistant"),
    UserMessage(content="What is Browser Use?")
]

response = await llm.ainvoke(messages)
print(response.completion)  # String response
print(response.usage)       # Token usage information

​
Parameters

  • messages (list[BaseMessage]): List of messages to send
  • output_format (type[T] | None): Optional Pydantic model for structured output
  • request_type (str): Type of request - "browser_agent" or "judge"
  • session_id (str | None): Session ID for sticky routing (same session → same container)

​
Returns

ChatInvokeCompletion[T] | ChatInvokeCompletion[str] with:
  • completion: Response content (string or structured output)
  • usage: Token usage information (prompt_tokens, completion_tokens, total_tokens)

​
Why ChatBrowserUse?

  1. Optimized for Browser Automation: Models fine-tuned specifically for browser tasks
  2. 3-5x Faster: Completes tasks significantly faster than generic models
  3. Lowest Cost: Most cost-effective solution for browser automation
  4. Built-in Retry Logic: Automatic handling of rate limits and transient errors
  5. Easy Setup: Simple API key configuration with $10 free credits