Amazon Bedrock

How to Deploy AI Agents with Amazon Bedrock AgentCore

Honghao Wang

— 2 min read

Amazon Bedrock AgentCore — Quick Start Guide

Amazon Bedrock AgentCore is a fully managed AWS service to build, deploy, and operate AI agents securely at scale. It integrates seamlessly with popular frameworks — Strands Agents, LangGraph, CrewAI, and LlamaIndex — and handles complex operational tasks such as runtime setup, IAM role configuration, and observability.

This guide walks you through:

  • Environment setup
  • Creating and testing a local agent
  • Deploying to AgentCore
  • Invoking via AWS SDK

---

📑 Table of Contents

---

Prerequisites

Ensure you have:

  • AWS account with CLI credentials
  • AWS CLI installed
  • Python 3.10+
  • `boto3` installed
  • Bedrock model access (e.g., Anthropic Claude Sonnet 4.0) enabled

---

Step 1 — AWS CLI Setup

Install the AWS CLI (docs).

Configure credentials:

aws configure

Or set up AWS SSO:

aws configure sso --profile my-profile

You will be prompted for:

  • SSO start URL
  • SSO region
  • AWS Account ID
  • Role name
  • Default region & output format

---

✅ Verify Authentication

aws sts get-caller-identity --profile my-profile

This returns:

  • Account ID
  • UserId (IAM User/Role)
  • ARN

If this works, you’re ready to use AWS SDK or Bedrock AgentCore.

---

Step 2 — Install & Create Agent

Virtual Environment

Create and activate a Python virtual environment:

macOS/Linux:

python3 -m venv .venv
source .venv/bin/activate

Windows:

python -m venv .venv
.venv\Scripts\activate

Deactivate with:

deactivate

---

Dependencies

Create `requirements.txt`:

bedrock-agentcore
strands-agents

Install:

pip install -r requirements.txt

---

Agent Script

Create `my_agent.py`:

from bedrock_agentcore import BedrockAgentCoreApp
from strands import Agent

app = BedrockAgentCoreApp()
agent = Agent()

@app.entrypoint
def invoke(payload):
    """AI agent function"""
    user_message = payload.get("prompt", "Hello! How can I help you today?")
    result = agent(user_message)
    return {"result": result.message}

if __name__ == "__main__":
    app.run()

---

Code Walkthrough

  • `BedrockAgentCoreApp()` → Sets up the agent runtime container
  • `Agent()` → Creates a basic Strands agent
  • `@app.entrypoint` → Entry point for requests
  • Reads `"prompt"` from payload (default message if not provided)
  • Returns agent’s output

---

Step 3 — Local Testing

Run:

python3 -u my_agent.py

In separate terminal:

curl -X POST http://localhost:8080/invocations \
-H "Content-Type: application/json" \
-d '{"prompt": "Hello!"}'

Expected:

{"result": "Hello! I'm here to help..."}

Stop with `Ctrl+C`.

---

Step 4 — Deploy to AgentCore

Configure:

agentcore configure -e my_agent.py

Creates `bedrock_agentcore.yaml`.

Launch:

agentcore launch

Output includes:

  • Agent ARN
  • CloudWatch Logs location

---

Test Deployed Agent

agentcore invoke '{"prompt": "tell me a joke"}'

If you get a joke back, deployment succeeded.

---

Step 5 — Invoke via AWS SDK

Create `invoke_agent.py`:

import json
import boto3

agent_arn = "YOUR_AGENT_ARN"
prompt = "Tell me a joke"

agent_core_client = boto3.client("bedrock-agentcore")
payload = json.dumps({"prompt": prompt}).encode()

response = agent_core_client.invoke_agent_runtime(
    agentRuntimeArn=agent_arn,
    payload=payload
)

content = []
for chunk in response.get("response", []):
    content.append(chunk.decode("utf-8"))
print(json.loads("".join(content)))

Run:

python invoke_agent.py

You should see the agent’s response.

---

Step 6 — Clean Up

Remove deployed runtime:

aws bedrock-agentcore delete-agent-runtime --agent-runtime-arn

---

Troubleshooting

  • Permission denied → Check AWS credentials/IAM policies
  • Docker warning → Ignore unless using `--local` build
  • Model access denied → Enable model in Bedrock console
  • Build errors → Check CloudWatch logs & IAM permissions

---

Conclusion

Amazon Bedrock AgentCore makes AI agent deployment fast and simple — no manual container setup needed. Prototype locally, deploy with one command, and monitor performance via CloudWatch.

For multi-platform publishing & monetization of your agent’s outputs, check AiToEarn — open-source AI content monetization platform supporting Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, and X (Twitter). Learn more at AiToEarn Blog or see AI Model Rankings.

---

References

Read more