This example demonstrates using AgentCore Browser with the Strands Agents framework for web automation and research tasks.

• AWS-managed browser - No custom browser configuration needed
• Strands integration - Uses strands_tools.browser.AgentCoreBrowser
• Streaming responses - Real-time output via agent.stream_async()
• Financial analysis - Example use case for stock data extraction

strands-browser/
├── serverless.yml      # Serverless configuration
├── agent.py            # Strands agent with browser tool
├── pyproject.toml      # Python dependencies
├── Dockerfile          # Container configuration
├── test-invoke.py      # Test script
└── README.md           # This file

serverless deploy

After deployment, note the runtime endpoint URL from the output (the ARN is embedded in the URL path):

ai:
  agents:
    browserAgent: https://bedrock-agentcore.us-east-1.amazonaws.com/runtimes/arn:aws:bedrock-agentcore:us-east-1:123456789012:runtime/xxx/invocations

export RUNTIME_ARN="arn:aws:bedrock-agentcore:us-east-1:123456789012:runtime/xxx"
python test-invoke.py

This example uses the AWS-managed default browser. No ai.browsers configuration is needed in serverless.yml:

ai:
  agents:
    browserAgent: {} # Auto-detects Dockerfile

The agent uses AgentCoreBrowser from strands_tools.browser:

from strands_tools.browser import AgentCoreBrowser

# Initialize browser (uses AWS-managed infrastructure)
browser_tool = AgentCoreBrowser(region="us-east-1")

# Create agent with browser capability (default: global.anthropic.claude-sonnet-5, override via MODEL_ID)
agent = Agent(
    tools=[browser_tool.browser],
    model=MODEL_ID
)

Web Search:

Search for the latest news about AWS and summarize the top 3 headlines

Financial Analysis:

Analyze the Tesla stock page at https://www.marketwatch.com/investing/stock/tsla
and provide key financial metrics

Data Extraction:

Visit https://aws.amazon.com/about-aws/whats-new/ and list the 5 most recent announcements

If you need session recording, VPC access, or request signing, define a custom browser:

ai:
  browsers:
    customBrowser:
      network:
        mode: PUBLIC
      recording:
        enabled: true
        s3Location:
          bucket: my-recordings
          prefix: sessions/

  agents:
    browserAgent: {}

See the Browser documentation for full configuration options.

serverless remove

• Browser Documentation - Full configuration reference
• Memory Example - Add conversation persistence
• Gateway Example - Add custom Lambda tools