Strands & AWS Bedrock AgentCore example

This repo is an example repo used in the blog: https://blog.elva-group.com/lets-build-an-agent-on-aws

Requirements

• Python 3.13
• uv
• AWS Credentials
• (optional) docker or podman for local container building

Getting started

1. Run uv sync
2. Activate venv via source .venv/bin/activate
3. Set your AWS Credentials as environment variables or similar
4. Run un run main.py to start the agent locally
5. Query the agent via curl or similar to get an streaming response:

   curl -X POST http://localhost:8080/invocations \
   -H "Content-Type: application/json" \
   -d '{"prompt": "Which tools do you have available?"}'
   

6. To skip the AgentCore framework, run the standalone.py directly to directly invoke the agent in your console. uv run standalone.py. Adjust the last line to play around with different questions like: agent("What is the best color?")

Deployment

1. Adjust account in .bedrock_agentcore.yaml
2. Make sure your AWS credentials is set and you are using a AgentCore supported region such as eu-west-1
3. run uv run agentcore launch to deploy the agent to AWS Bedrock AgentCore, after the deployment you can find your agent in the console