Aws Strands

The Strands Agents SDK gives you a model-agnostic framework for building AI agents that works with Claude, GPT-4, or any LLM you point it at. It's built for the ReAct pattern out of the box and handles multi-agent orchestration when you need specialist agents working together. The decorator-based tool definition is clean, and it integrates with AWS Bedrock AgentCore if you're already in that ecosystem. AWS features it in their official AgentCore samples for multi-agent incident response. If you're tired of vendor lock-in or need to swap models without rewriting your agent logic, this is worth looking at.

skills.sh installs: 249
GitHub stars: 296
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add hoodini/ai-agents-skills --skill aws-strands --agent claude-code

Installs into .claude/skills of the current project.

Files

Strands Agents SDK

The open-source framework you write an agent's logic in - the "brain." Model-agnostic, AWS Bedrock by default. Runs anywhere (laptop, container, Lambda, EC2).

How this fits with the other AWS skill: Strands is the FRAMEWORK (what your agent does). To HOST and DEPLOY a Strands agent on AWS, use aws-harness (the AgentCore runtime). They compose: write with Strands, ship with AgentCore. You can also run Strands with no AWS deployment at all.

Install

pip install strands-agents strands-agents-tools

(A TypeScript SDK also exists - see the docs. Examples below are Python.)

Quick Start

from strands import Agent

agent = Agent()                       # defaults to Bedrock, Claude 4 Sonnet
print(agent("What is the capital of France?"))

agent(...) returns an AgentResult. str(result) gives the text; result.message is the structured dict (role + content).

Model configuration

Bedrock is the default provider and no model argument is needed - Strands picks a region-appropriate Claude 4 Sonnet. To override, pass a model id string or a BedrockModel provider:

SKILL.md on GitHub