Strands Agents
AWS's open-source (Apache 2.0) SDK for building AI agents with a model-driven approach - an agent in just a few lines of code. Python and TypeScript, any model, any cloud.
Description
Strands Agents is an open-source (Apache 2.0) toolkit (SDK) created by AWS / Amazon for building production AI agents. Its core idea is a model-driven approach: instead of hand-coding complex control flow, a developer defines a model, a set of tools and a prompt, and the agent plans its own next steps and calls tools. As a result a simple agent can be built in just a few lines of code.
The SDK is available in two languages: Python (pip install strands-agents) and TypeScript (npm install @strands-agents/sdk). Strands is model-provider agnostic ('any model, any cloud') - it supports Amazon Bedrock, Anthropic, OpenAI, Llama, Ollama and LiteLLM among others, letting you swap the model backend as you scale without rewriting the agent.
Its main mechanisms include: defining tools via decorators and having the agent invoke them autonomously, support for the Model Context Protocol (MCP) to ground agents in external knowledge sources and tools, a hooks system to intercept and modify agent behavior, conversation context management (summarization, sliding window), guardrails and policy enforcement before tool use, structured output (Pydantic / Zod), and built-in observability (traces, OpenTelemetry).
Strands supports multi-agent patterns (Agent-as-Tool, Swarm) and a broad range of deployment options: AWS Lambda (serverless), Fargate, EKS, EC2/AgentCore, as well as Docker and Terraform - an agent can run locally, in managed cloud or at the edge. The project was open-sourced in May 2025 and has gained over 6,700 GitHub stars.