Browserbase

BrowserbaseTools enable an Agent to automate browser interactions using Browserbase, a headless browser service.

Prerequisites

The following example requires Browserbase API credentials after you signup here, and the Playwright library.

uv pip install agno browserbase playwright openai
export BROWSERBASE_API_KEY=xxx
export BROWSERBASE_PROJECT_ID=xxx

Example

The following agent will use Browserbase to visit https://quotes.toscrape.com and extract content. Then navigate to page two of the website and get quotes from there as well.

cookbook/91_tools/browserbase_tools.py

from agno.agent import Agent
from agno.tools.browserbase import BrowserbaseTools

agent = Agent(
    name="Web Automation Assistant",
    tools=[BrowserbaseTools()],
    instructions=[
        "You are a web automation assistant that can help with:",
        "1. Capturing screenshots of websites",
        "2. Extracting content from web pages",
        "3. Monitoring website changes",
        "4. Taking visual snapshots of responsive layouts",
        "5. Automated web testing and verification",
    ],
    markdown=True,
)

agent.print_response("""
    Visit https://quotes.toscrape.com and:
    1. Extract the first 5 quotes and their authors
    2. Navigate to page 2
    3. Extract the first 5 quotes from page 2
""")

See the Startup Analyst MCP agent

Close the browser connection

Retain the toolkit instance and close it even when the agent run fails:

from agno.agent import Agent
from agno.tools.browserbase import BrowserbaseTools

browser = BrowserbaseTools()
try:
    Agent(tools=[browser]).print_response(
        "Read the first five quotes from https://quotes.toscrape.com."
    )
finally:
    browser.close_session()

For an async run, use await browser.aclose_session() in an async finally block. These methods disconnect local Playwright resources. Normal provider sessions end on disconnect; an externally supplied keep-alive session needs the provider's REQUEST_RELEASE operation to end immediately.

Toolkit Params

Parameter	Type	Default	Description
`api_key`	`str`	`None`	Browserbase API key. If not provided, uses BROWSERBASE_API_KEY env var.
`project_id`	`str`	`None`	Browserbase project ID. If not provided, uses BROWSERBASE_PROJECT_ID env var.
`base_url`	`str`	`None`	Custom Browserbase API endpoint URL. This changes the HTTP API endpoint; it does not select a browser session region. If not provided, uses BROWSERBASE_BASE_URL env var.
`enable_navigate_to`	`bool`	`True`	Enable the navigate_to functionality.
`enable_screenshot`	`bool`	`True`	Enable the screenshot functionality.
`enable_get_page_content`	`bool`	`True`	Enable the get_page_content functionality.
`enable_close_session`	`bool`	`True`	Enable the close_session functionality.
`all`	`bool`	`False`	Enable all functionality.
`parse_html`	`bool`	`True`	Strip script/style blocks and HTML tags with regular expressions; this does not evaluate CSS visibility. Reduces token usage.
`max_content_length`	`Optional[int]`	`100000`	Maximum character length for page content. Longer content is truncated with a notice. Set to `None` for no limit.

Toolkit Functions

Function	Description
`navigate_to`	Navigates to a URL. Takes a URL and an optional connect_url parameter.
`screenshot`	Takes a screenshot of the current page. Takes a path to save the screenshot, a boolean for full-page capture, and an optional connect_url parameter.
`get_page_content`	Gets the content of the current page. Returns text-only content when `parse_html=True` (default), otherwise raw HTML. Takes an optional connect_url parameter.
`close_session`	Closes a browser session.

Developer Resources

Tools