Strands Agents — Open Source AI Agent SDK Documentation
by AWS (Strands Agents)
AWS's model-driven agent SDK, documented end to end — agent loop, tools, MCP, multi-agent, sessions, and deployment on Lambda, Fargate, EKS, or Docker.
Overview
Strands Agents bills itself as 'the open source toolkit for building production agents', and its documentation site is the reference for both the Python SDK (pip install strands-agents) and the TypeScript SDK (npm install @strands-agents/sdk). The approach is model-driven: rather than hand-coding an orchestration graph, you give the model a prompt and a set of tools and let its reasoning loop drive execution — the docs open by defining an agent with a tool the model can call during that loop. The framework was developed at AWS from systems already running in production inside Amazon (Amazon Q Developer, AWS Glue, and VPC Reachability Analyzer among them) and is deliberately model- and cloud-agnostic: 'any model, any cloud', with support spanning Amazon Bedrock, Anthropic's Claude, and other providers. The documentation covers the agent loop fundamentals, creating and managing tools, Model Context Protocol (MCP) integration, multi-agent orchestration patterns, conversation and session management (including retrieving agent state from a remote datastore), built-in observability and tracing, and deployment guidance across AgentCore, AWS Lambda, Fargate, EKS, and Docker. Version 1.0 added four multi-agent primitives, support for the Agent2Agent (A2A) protocol, a session manager, and improved async support throughout the SDK. The project is open source (6,500+ GitHub stars) and free to use.
At a Glance
- Topic
- Frameworks
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Self-paced reference; quickstart in under an hour
- Provider
- AWS (Strands Agents)
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Define a working agent in a few lines of Python or TypeScript using the model-driven agent loop
- ✓Write, register, and manage custom tools the model can call during its reasoning loop
- ✓Connect agents to MCP servers to consume external tools and context
- ✓Compose multi-agent systems using Strands' orchestration primitives and the Agent2Agent (A2A) protocol
- ✓Manage conversation state and sessions, including persisting agent state to a remote datastore
- ✓Instrument agents with built-in observability and tracing for production debugging
- ✓Deploy agents to AgentCore, AWS Lambda, Fargate, EKS, or Docker
Highlights
- •Built by AWS from agents already running in production inside Amazon, not a demo framework
- •Model- and cloud-agnostic by design — works with Bedrock, Claude, and third-party providers
- •First-class Python and TypeScript SDKs, not a Python library with a bolted-on port
- •Deployment and observability are documented as first-class concerns, not afterthoughts
Who It’s For
Best For
- ✓AI engineers who want a production agent framework with minimal boilerplate
- ✓AWS-centric teams building agents on Bedrock, Lambda, Fargate, or EKS
- ✓TypeScript developers looking for a serious agent SDK outside the Python ecosystem
- ✓Teams comparing agent SDKs (OpenAI Agents SDK, LangGraph, Microsoft Agent Framework, Strands)
Prerequisites
- •Comfortable with Python or TypeScript
- •Familiarity with LLM tool/function calling
- •Basic AWS knowledge helps for the deployment chapters but is not required to use the SDK
FAQ
What is Strands Agents — Open Source AI Agent SDK Documentation?
The official documentation for Strands Agents, the open-source, model-driven agent SDK that AWS built from production systems inside Amazon and released for Python and TypeScript. It is for engineers who want to define an agent in a few lines of code and still have a credible path to production observability and deployment.
Is Strands Agents — Open Source AI Agent SDK Documentation free?
Strands Agents — Open Source AI Agent SDK Documentation is free to access.
What level is Strands Agents — Open Source AI Agent SDK Documentation for?
Strands Agents — Open Source AI Agent SDK Documentation is aimed at a intermediate audience. Recommended background: Comfortable with Python or TypeScript, Familiarity with LLM tool/function calling, Basic AWS knowledge helps for the deployment chapters but is not required to use the SDK.
How long does Strands Agents — Open Source AI Agent SDK Documentation take?
Expect roughly Self-paced reference; quickstart in under an hour. Most learners work through it at their own pace.
What will I learn from Strands Agents — Open Source AI Agent SDK Documentation?
You'll learn: Define a working agent in a few lines of Python or TypeScript using the model-driven agent loop; Write, register, and manage custom tools the model can call during its reasoning loop; Connect agents to MCP servers to consume external tools and context; Compose multi-agent systems using Strands' orchestration primitives and the Agent2Agent (A2A) protocol; Manage conversation state and sessions, including persisting agent state to a remote datastore; Instrument agents with built-in observability and tracing for production debugging; Deploy agents to AgentCore, AWS Lambda, Fargate, EKS, or Docker.