Enhancing AI Agent Capabilities with AgentCore Browser: New Features for Seamless Web Interactions

Unlocking Advanced Browser Features for AI Agents

Persistent Browser Profiles: Keeping Your AI Agents Upto Speed

Secure Traffic Management with Proxy Configuration

Customizable Browser Behavior through Extensions

Conclusion: Empowering AI Agents for Real-World Applications

About the Authors: Meet the Innovators Behind AgentCore Browser

Unlocking the Future of AI with Advanced AgentCore Browser Features

In an age where AI is transforming industries, the capabilities of AI agents tasked with web browsing have evolved beyond basic page navigation. Our customers have voiced their needs for agents that can retain session states across interactions, route traffic through corporate proxies, and adapt to custom browser configurations. Enter AgentCore Browser—a secure, isolated environment designed for AI agents engaging with web applications.

Previously, each session in AgentCore Browser commenced with a clean slate, limited by default settings and direct internet access. This often hampered the real-world functionality needed in enterprise environments. Today, we are thrilled to announce three game-changing capabilities: proxy configuration, browser profiles, and browser extensions. These additions empower you with fine-grained control over AI agents’ web interactions.

The New Capabilities Explained

With these enhancements, you’ll be able to dictate how your AgentCore Browser sessions connect to the internet, what states they retain, and how they operate. Let’s delve into each feature:

1. Proxy Configuration

Our newfound support for proxy configurations allows browser traffic to pass through your external proxy servers. This spares your agents from unwanted interruptions while ensuring that traffic adheres to corporate network infrastructure.

How It Works

To initiate a browser session with proxy settings, use the StartBrowserSession method alongside a specified proxy configuration. If needed, you can also store your proxy credentials securely using AWS Secrets Manager. Here’s a simplified example:

import boto3
import json
client = boto3.client('secretsmanager')
client.create_secret(
    Name="my-proxy-credentials",
    SecretString=json.dumps({
        'username': 'your_username',
        'password': 'your_password'
    })
)

2. Browser Profiles for Session Continuity

Browser profiles are critical for maintaining an agent’s context across sessions. Imagine an agent engaged in e-commerce testing or multi-step user journeys that require re-authentication at every turn. With browser profiles, your agents will save and restore cookies and local storage, allowing them to pick up right where they left off.

Example Usage

Here is how you can create and use browser profiles:

# Create a browser profile
control_client = boto3.client('bedrock-agentcore-control')
profile = control_client.create_browser_profile(name="ecommerce_profile")

# Start a browser session, add items, and save the session
session1 = session_client.start_browser_session(browserIdentifier="aws.browser.v1")
# ... agent navigates and adds items to cart ...
session_client.save_browser_session_profile(sessionId=session1['sessionId'], browserIdentifier="aws.browser.v1", profileIdentifier=profile['profileId'])

3. Custom Browser Extensions

Adding browser extensions enables you to fine-tune your agents’ browsing behavior significantly. Whether you need ad-blocking, authentication helpers, or other tailored functionalities, packaging your extension as a ZIP file and loading it into AgentCore Browser takes mere moments.

Starting a Session with Extensions

Once your extension is uploaded to Amazon S3, initiate a browser session that leverages these tools like so:

s3.upload_file('my-extension.zip', 'your-bucket-name', 'extensions/my-extension.zip')

response = client.start_browser_session(
    browserIdentifier="aws.browser.v1",
    name="session-with-extensions",
    extensions=[{"location": {"s3": {"bucket": "your-bucket-name", "prefix": "extensions/my-extension.zip"}}}]
)

Conclusion

With new capabilities in proxy configuration, browser profiles, and browser extensions, the AgentCore Browser is set to revolutionize how AI agents interact with the web. These enhancements facilitate seamless traffic routing, ensure session continuity, and allow custom browser behavior—all while safeguarding credentials using AWS Secrets Manager.

To embark on this journey, check out the tutorials in the Amazon Bedrock AgentCore samples repository and explore the extensive documentation. For pricing details, visit the Amazon Bedrock AgentCore Pricing page.

About the Authors

Joshua Samuel – Senior AI/ML Specialist Solutions Architect

Based in Melbourne, Australia, Joshua specializes in agentic AI and coding techniques to accelerate enterprise transformation.

Evandro Franco – Sr. Data Scientist

Part of the AWS Global GTM team, Evandro helps clients tackle AI/ML challenges leveraging Amazon Bedrock solutions.

Kosti Vasilakakis – Principal PM

Kosti has a wealth of experience in building and designing Bedrock AgentCore services, drawing from his background as a data scientist.

Yan Marim – Sr. GenAI Specialist Solutions Architect

Focusing on generative AI adoption in Brazil, Yan helps guide customers through their transition to AI solutions.

Kevin Orellana – Software Development Engineer

Based in Seattle, Kevin operates core infrastructure supporting agentic AI capabilities, continually pushing the boundaries in software development.

These advancements symbolize a significant leap for organizations looking to deploy AI agents that navigate the web effortlessly and intelligently. Embrace these innovations to enhance your enterprise capabilities today!