Strands Agents is an open-source agent SDK from AWS, built from systems that run inside Amazon and open sourced in 2025. It treats a model-driven loop as the whole framework: an agent plans with the tools you define, scaling from a single function call to a multi-agent team, on any model and any cloud.
AWS describes it as “built from production systems inside Amazon” (Strands Agents), open sourced after running internal workloads. Since May 2025 it has reached about 37 million PyPI downloads per month as of August 2026, within reach of far older frameworks.
How is Strands different from other agent frameworks?
Three things. It is minimal by default: an AI agent is a model plus tools plus a loop, and the framework adds structure only when you ask for it. It is model-agnostic: unlike vendor-native SDKs, it does not assume a specific provider. It comes from production code rather than a paper: the abstractions survived real Amazon workloads before they faced the public, which tends to produce opinionated defaults around tool calling and error handling instead of around demos.
When does it earn a place in a stack?
When your infrastructure already lives on AWS and you want an agent layer that integrates there without locking your model choice. The SDK scales from one tool to agent teams without a rewrite. The caveat is community maturity: documentation is strong, but the ecosystem of examples and battle stories is younger than LangChain’s or CrewAI’s, so expect problems nobody has published a solution to yet.
Frequently asked questions
What does it mean that Strands Agents is model-agnostic?
An agent built with Strands runs against any model provider; AWS describes it as working with any model on any cloud. The SDK ships from AWS but does not lock the underlying model, which distinguishes it from vendor-native SDKs such as the OpenAI Agents SDK.
Why do Strands Agents downloads look high compared with its GitHub stars?
The package is wired into AWS tooling, examples, and training material, so automated installs inflate the count relative to human awareness measured by stars. Read the download number as adoption pressure inside the AWS ecosystem rather than as a community-size measure.
No advertising or tracking cookies, and our visitor counts are anonymous. The Cal.com booking widget loads only if you allow it. Privacy Policy.
The page itself, anything our host sets to serve and secure it, and the anonymous visitor count. Always on, and none of it stores anything on your device.
The Cal.com booking widget. Left off, a booking link opens the booking page instead of a popup, so you can still book a call.