Getting Started with Amazon Bedrock AgentCore Browser Tool

9 min readSep 28, 2026

Introduction

AI agents become significantly more useful when they can browse the web. Instead of relying solely on a model’s training data, you can give your agent the ability to navigate to specific URLs, interact with page elements, and extract content from web pages in real time.

The Amazon Bedrock AgentCore Browser Tool provides a managed, secure, cloud-based browser runtime on AWS infrastructure. When an agent needs to access a web page, it opens a remote browser session, navigates to the target URL, and interacts with the page just as a human would. The session is isolated and automatically cleaned up when done.

This article is part of a series I am writing about different aspects of Amazon Bedrock AgentCore. Here, we focus on how to use the Browser Tool to give agents the ability to navigate and extract content from web pages. For a broader understanding of AgentCore and its components, check out the other articles in this series:

In this article, we explore the Browser Tool from two perspectives:

First, we invoke the tool locally using the Strands Agents SDK, demonstrating how developers can use it in isolation outside of a Harness.
Then, we connect the tool to the car dealership Harness introduced in the first article and let the AI agent use it autonomously.

What is the Browser Tool?

The AgentCore Browser Tool provides a managed, secure, cloud-based browser runtime on AWS infrastructure. Unlike the Web Search Tool (covered in the previous article), which searches the web and returns results, the Browser Tool navigates to a specific URL and interacts with the page directly.

Key capabilities:

Navigate to any URL
Read page content and extract text
Click elements, fill forms, and interact with the page
Take screenshots
Handle authentication flows
Session recording for debugging

You can use the system default browser or create a custom Browser Tool with specific configurations (such as a browser profile for persisting cookies across sessions).

Use the Browser Tool when you know the URL and need to extract specific content, fill forms, or interact with web applications.

Prerequisites

AWS account with credentials configured
Python 3.12 installed (Python 3.14 has known compatibility issues with Playwright)
IAM user or role with permissions to use Amazon Bedrock AgentCore Browser and invoke Bedrock models
Model access enabled for Amazon Nova Pro in the Amazon Bedrock console
AWS Region where AgentCore is available (us-east-1, us-west-2, ap-southeast-2, or eu-central-1)

Step 1: Clone the Repository and Install Dependencies

Clone the companion repository and set up the Python environment:

git clone https://github.com/biagolini/PythonAwsBedrockAgentCoreBrowserTool.git
cd PythonAwsBedrockAgentCoreBrowserTool
python3.12 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

On Windows, replace the activation command with .venv\Scripts\activate.

Step 2: Create the Browser Tool Execution Role

Before creating the Browser Tool, you need an IAM role that allows the AgentCore Browser service to access resources on your behalf (such as writing session recordings to S3).

2.1. Navigate to IAM > Roles in the AWS Console

2.2. Click Create role

2.3. Select Custom trust policy and enter the following trust policy:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "AssumeRolePolicy",
      "Effect": "Allow",
      "Principal": {
        "Service": "bedrock-agentcore.amazonaws.com"
      },
      "Action": "sts:AssumeRole",
      "Condition": {
        "StringEquals": {
          "aws:SourceAccount": "<your-account-id>"
        },
        "ArnLike": {
          "aws:SourceArn": "arn:aws:bedrock-agentcore:<your-region>:<your-account-id>:*"
        }
      }
    }
  ]
}

2.4. Click Next. Skip managed policies for now. Click Next.

2.5. In the Role name field, enter AgentCoreBrowser-S3-Recording

2.6. Click Create role

2.7. After creation, open the role and click Add permissions > Create inline policy. Select the JSON tab and enter the following policy:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "S3Permissions",
      "Effect": "Allow",
      "Action": [
        "s3:PutObject",
        "s3:GetObject",
        "s3:ListBucket",
        "s3:ListMultipartUploadParts",
        "s3:AbortMultipartUpload"
      ],
      "Resource": [
        "arn:aws:s3:::<your-recording-bucket-name>",
        "arn:aws:s3:::<your-recording-bucket-name>/*"
      ]
    },
    {
      "Sid": "CloudWatchLogsPermissions",
      "Effect": "Allow",
      "Action": [
        "logs:CreateLogGroup",
        "logs:CreateLogStream",
        "logs:PutLogEvents",
        "logs:DescribeLogStreams"
      ],
      "Resource": "*"
    }
  ]
}

Replace <your-recording-bucket-name> with the S3 bucket name you plan to use for session recordings. If you do not plan to enable recording, you can omit the S3 permissions and keep only the CloudWatch Logs permissions.

2.8. Click Next, name the policy AgentCoreBrowser-S3-RecordingPolicy, and click Create policy

Step 3: Create a Browser Tool in the Console

Before your agent can browse the web, you need to create a Browser Tool resource. You can use the system default browser (no setup required) or create a custom one for additional control.

3.1 Navigate to Browser Tool

3.1.1. Open the Amazon Bedrock AgentCore console

3.1.2. In the left navigation pane, under Built-in tools, click Browser

3.1.3. Click the Create button

Browser Tool console showing the navigation path and Create button

3.2 Configure the Browser Tool

3.2.1. In the Browser name field, enter a name for your browser (example: browser_tool_demo). Valid characters are a-z, A-Z, 0-9, hyphen, and underscore.

3.2.2. Optionally, configure a recording S3 bucket to save session recordings for debugging. This is useful during development to see exactly what the browser did during a session.

3.2.3. Under Permissions, select Use an existing role and choose the AgentCoreBrowser-S3-Recording role created in Step 2.

3.2.4. Leave remaining settings at their defaults.

Get Carlos Biagolini’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

3.2.5. Click Create

Create Browser Tool form showing name, recording, and permissions configuration

3.3 Copy the Browser Tool ID

After creation, the console displays the browser details page. Copy the Browser ID value. You will use this identifier in all API calls.

Browser Tool details page showing the Browser ID

Step 4: Test the Browser Tool Locally

The script browser_agent.py demonstrates how to use the Browser Tool with the Strands Agents SDK. It creates an agent that navigates to the AgentCore documentation page and extracts a summary.

4.1 Configure Environment Variables

Copy the example environment file and add your Browser Tool ID:

cp .env.example .env

Edit .env:

AWS_REGION=us-east-1
BROWSER_ID=<your-browser-tool-id>

4.2 Understanding the Script

import os
import sys
import logging
import warnings
from dotenv import load_dotenv

# Suppress asyncio and Playwright noise
logging.getLogger("asyncio").setLevel(logging.CRITICAL)
warnings.filterwarnings("ignore", category=DeprecationWarning)

load_dotenv()

REGION = os.getenv("AWS_REGION", "us-east-1")
BROWSER_ID = os.getenv("BROWSER_ID")

if not BROWSER_ID:
    print("Error: BROWSER_ID not set in .env file")
    sys.exit(1)

from strands import Agent
from strands.models.bedrock import BedrockModel
from strands_tools.browser import AgentCoreBrowser

browser_tool = AgentCoreBrowser(region=REGION, identifier=BROWSER_ID)

model = BedrockModel(
    model_id="us.amazon.nova-pro-v1:0",
    region_name=REGION,
)

agent = Agent(
    model=model,
    tools=[browser_tool.browser],
    system_prompt=(
        "You are a research assistant. You MUST use the browser tool to navigate "
        "to URLs and read their content. Never answer from memory — always use "
        "the browser tool to fetch the page content first, then summarize what you read."
    ),
)

prompt = (
    "Use the browser tool to go to "
    "https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html "
    "and based on what you read on that page, provide a short explanation (3-4 sentences) "
    "of what Amazon Bedrock AgentCore is and its main purpose."
)

print(f"Prompt: {prompt}\n")

try:
    response = agent(prompt)
    print("\n--- Agent Response ---")
    print(response.message["content"][0]["text"])
except Exception as e:
    print(f"\nError: {e}")
finally:
    print("\nDone.")
    os._exit(0)

The key components are:

AgentCoreBrowser from strands_tools.browser — initializes the browser tool client with your custom Browser ID
The agent’s system prompt instructs it to always use the browser tool rather than answering from memory
The prompt asks the agent to navigate to a specific URL and extract information

Note that we target an HTML documentation page rather than a PDF. Complex SPAs with heavy JavaScript and asynchronous rendering can sometimes cause issues with current library versions. Static or server-rendered pages provide the most reliable results. These limitations are expected to improve in future versions.

4.3 Run the Script

python browser_agent.py

Expected behavior:

Prompt: Use the browser tool to go to https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html and based on what you read on that page, provide a short explanation (3-4 sentences) of what Amazon Bedrock AgentCore is and its main purpose.

Tool #1: browser
Tool #2: browser
Tool #3: browser
...
--- Agent Response ---
Amazon Bedrock AgentCore is an agentic platform that enables developers to build, deploy, and operate highly effective agents securely at scale using any framework and foundation model. Its main purpose is to allow agents to take actions across tools and data with the right permissions and governance, run agents securely at scale, and monitor agent performance and quality in production without requiring infrastructure management.
Done.

The agent invokes the browser tool multiple times (session initialization, navigation, content extraction) and returns an accurate summary based on the actual page content. The process exits cleanly after printing the response.

Step 5: Browser Sessions and Observability

After running the script, you can verify the browser session in the AgentCore console.

5.1 View Sessions

Navigate to Amazon Bedrock AgentCore > Built-in tools > Browser, select your browser, and check the Sessions tab. You will see the session that was created during the test, along with its status and duration.

Browser Tool sessions list in the console

5.2 Live View (Optional)

If you configured session recording, you can also access a live view of what the browser is doing during an active session. This is invaluable for debugging complex navigation scenarios.

Step 6: Integrating with the Harness

Now we connect the Browser Tool to the Harness so the agent can use it autonomously.

6.1 Edit the Harness to Add the Browser Tool

6.1.1. Navigate to Amazon Bedrock AgentCore > Harness

6.1.2. Click on your Harness name (car_agent) to open its details page

6.1.3. Click Edit in the top-right area

6.1.4. In the Tools section, enable Browser tool and select your custom browser (browser_tool_demo) from the dropdown

Harness test showing Browser Tool navigation and content extraction

6.1.5. Click Save Harness and wait for the status to return to Ready

6.2 Test the Browser Tool via Harness

6.2.1. Once the Harness is Ready, click the Test Harness button

6.2.2. Before testing, check if the Session ID field already has a value from a previous test. If so, click New session and then Stop current session and create new session to start fresh. This avoids interference from previous conversation context.

6.2.3. Enter a prompt that requires URL navigation:

Use the browser to go to https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/what-is-bedrock-agentcore.html and based on what you read on that page, provide a short explanation of what Amazon Bedrock AgentCore is.

6.2.4. The agent will invoke the browser tool, navigate to the URL, read the page content, and return a summary. You should see indicators confirming the tool was used.

6.2.5. Optionally, try editing the Harness to use a different model and repeat the test. Different models may produce different results in terms of speed and accuracy.

Conclusion

In this article, you learned how to use the Amazon Bedrock AgentCore Browser Tool to give agents the ability to navigate and interact with web pages:

The Browser Tool provides a managed, secure browser runtime that agents control to navigate URLs, interact with pages, and extract content. You created a custom Browser Tool, configured the Strands Agents SDK to use it, and demonstrated how the agent navigates to a documentation page and summarizes its content.
Local invocation lets developers test the tool in isolation using the Strands Agents SDK with AgentCoreBrowser.
Harness integration allows the agent to use the Browser Tool autonomously, deciding when to browse based on the conversation context.

The key takeaways are:

Browser Tool is for targeted navigation. Use it when you know the URL and need to extract specific content, fill forms, or interact with web applications.
Static pages work best. Complex SPAs with heavy JavaScript may cause issues with current versions. This is expected to improve over time.
Model choice matters. Larger models produce more reliable tool invocations with fewer retries.
Combine with Web Search. In a production agent, use the Web Search Tool to discover relevant pages, and the Browser Tool to navigate to those pages and extract detailed content.

References

Stay Connected

If you found this article helpful and want to learn more about AWS, cloud architecture, AI, infrastructure as code, and cloud security, follow the author for future content and tutorials:

LinkedIn: https://www.linkedin.com/in/biagolini
Medium: https://medium.com/@biagolini
GitHub: https://github.com/biagolini
YouTube (PT-BR): https://www.youtube.com/@BiagoliniTech
AWS Builder Center (PT-BR): https://builder.aws.com/community/@cbiagolini

Stay tuned for more practical walkthroughs and solutions for cloud infrastructure and DevOps challenges.

Happy building on AWS! 🚀☁️

AWS Tip

Getting Started with Amazon Bedrock AgentCore Browser Tool

Introduction

What is the Browser Tool?

Prerequisites

Step 1: Clone the Repository and Install Dependencies

Step 2: Create the Browser Tool Execution Role

Step 3: Create a Browser Tool in the Console

3.1 Navigate to Browser Tool

3.2 Configure the Browser Tool

Get Carlos Biagolini’s stories in your inbox

3.3 Copy the Browser Tool ID

Step 4: Test the Browser Tool Locally

4.1 Configure Environment Variables

4.2 Understanding the Script

4.3 Run the Script

Step 5: Browser Sessions and Observability

5.1 View Sessions

5.2 Live View (Optional)

Step 6: Integrating with the Harness

6.1 Edit the Harness to Add the Browser Tool

6.2 Test the Browser Tool via Harness

Conclusion

References

Stay Connected

Written by Carlos Biagolini

Published in AWS Tip

No responses yet

More from Carlos Biagolini and AWS Tip

Getting Started with Amazon Bedrock AgentCore Knowledge Bases

Introduction

Google Gave Me 45 Minutes to Design a URL Shortener. I Used 40 Minutes on the Wrong Thing.

Every engineer overthinks this question in exactly the similar way.

20 Terraform Best Practices I Wish I Had Learned Earlier

Hard-earned lessons to keep your infrastructure predictable, secure, and drama-free

Getting Started with Amazon CloudWatch for Bedrock AgentCore Observability

Introduction

Recommended from Medium

How To Run Claude Like a Tech Lead, Not a Senior Engineer

You already know how to brief one session. The leverage now is running four of them and still shipping code you’d put your own name on.

8 New AI Terms You Should Know to Be Ahead of 90% of People

A few more AI terms to know beyond tokens, prompts, and hallucinations.

Building an Agentic Knowledge Graph for Billions of Events

Rebuilding Netflix-inspired observability with AI

Infrastructure Documentation Best Practices: ADRs, Runbooks and Diagrams as Code

Infrastructure documentation has a reputation problem. It is written during calm periods, never updated during incidents, and discovered by…

How to Master API Load Testing with Grafana k6 and Visualize Real-Time Performance Metrics Like an…

Load testing ensures your applications survive traffic surges, but understanding performance metrics often feels overwhelming for…

Getting Started with Amazon Bedrock AgentCore Identity

Introduction

No responses yet

More from Carlos Biagolini and AWS Tip

Getting Started with Amazon Bedrock AgentCore Knowledge Bases

Introduction

Google Gave Me 45 Minutes to Design a URL Shortener. I Used 40 Minutes on the Wrong Thing.

Every engineer overthinks this question in exactly the similar way.

20 Terraform Best Practices I Wish I Had Learned Earlier

Hard-earned lessons to keep your infrastructure predictable, secure, and drama-free

Getting Started with Amazon CloudWatch for Bedrock AgentCore Observability

Introduction

Recommended from Medium

How To Run Claude Like a Tech Lead, Not a Senior Engineer

You already know how to brief one session. The leverage now is running four of them and still shipping code you’d put your own name on.

8 New AI Terms You Should Know to Be Ahead of 90% of People

A few more AI terms to know beyond tokens, prompts, and hallucinations.

Building an Agentic Knowledge Graph for Billions of Events

Rebuilding Netflix-inspired observability with AI

Infrastructure Documentation Best Practices: ADRs, Runbooks and Diagrams as Code

Infrastructure documentation has a reputation problem. It is written during calm periods, never updated during incidents, and discovered by…

How to Master API Load Testing with Grafana k6 and Visualize Real-Time Performance Metrics Like an…

Load testing ensures your applications survive traffic surges, but understanding performance metrics often feels overwhelming for…

Getting Started with Amazon Bedrock AgentCore Identity

Introduction