Strands Agents is an open-source SDK, maintained under the strands-agents GitHub organization with an Amazon.com copyright in its NOTICE file, for building and running AI agents in Python and TypeScript that run in-process with no hosted control plane. The monorepo ships two ways to use it: the Strands Agents SDK (strands-py on PyPI, strands-ts on npm), a lower-level agent loop where developers wire up tools, model providers, and memory themselves, invoked with a few lines such as Agent(tools=[calculator]); and Strands harness (strands-harness on PyPI, @strands-agents/harness on npm), a batteries-included agent assembled via a single create_harness() or createHarness() call with benchmarked defaults for the model, tools, memory, sessions, and context management. Both layers share the same core capabilities: lifecycle controls such as turn limits, token budgets, cancellation, and stop reasons; tool calling and structured output; MCP tool support; multi-agent patterns; memory and session management; streaming; guardrails meant to catch mistakes before they run and steering handlers that let agents correct themselves instead of failing silently; and built-in tracing and evals. Model support is provider-agnostic, with first-class integrations for Amazon Bedrock, Anthropic, OpenAI, and Gemini plus additional and custom providers documented on the project's site. The repository also includes a strands CLI (distributed on npm) for prototyping and chatting with a harness agent from the terminal, an MCP server package (strands-mcp), and the documentation site source. The Python SDK requires Python 3.10+; the TypeScript SDK requires Node.js 22+.