A model-driven, open source SDK for creating AI agents—from simple conversational assistants to complex autonomous workflows. Python and TypeScript. Apache 2.0.
from strands import Agent
from strands_tools import calculator
agent = Agent(tools=[calculator])
agent("What is the square root of 1764?")import { Agent } from "@strands-agents/sdk";
const agent = new Agent();
await agent.invoke("What is the square root of 1764?");pip install strands-agents strands-agents-tools # Python
npm install @strands-agents/sdk # TypeScript- Simple — A minimal agent loop that just works. No boilerplate, no complex abstractions.
- Model agnostic — Amazon Bedrock, Anthropic, OpenAI, Gemini, Ollama, LiteLLM, llama.cpp, and more. Swap providers in one line.
- Multi-agent built in — Coordinate agents with Graph, Swarm, and Workflow patterns — plus A2A protocol for cross-framework interop.
- Native MCP — First-class Model Context Protocol support gives your agents access to thousands of tools.
- Production ready — Deploy to AWS Lambda, Fargate, EKS, Bedrock AgentCore, Docker, Kubernetes, or Terraform with built-in OpenTelemetry observability.
- Conversational & beyond — Streaming, non-streaming, autonomous agents, structured output, and real-time bidirectional audio conversations.
Strands is built in the open and growing fast. The community contributes model providers (Cohere, xAI, Fireworks AI, NVIDIA NIM, vLLM, MLX, SGLang), session managers (Valkey/Redis, AgentCore Memory), tools (Telegram, HubSpot, Deepgram), and integrations (AG-UI).
👉 See all community packages · Get your project featured · Join the discussion
| sdk-python | Python SDK — the core agent framework |
| sdk-typescript | TypeScript SDK — agents in Node.js and the browser |
| samples | Example agents and real-world use cases |
| docs | Documentation source for strandsagents.com |
| mcp-server | Documentation server for building with Strands in your AI coding assistant |
| tools | Pre-built tools: file I/O, HTTP, shell, code interpreter, and more |
| evals | SDK for testing and benchmarking your agents |
| agent-sop | Structured prompting for reliable, multi-step agent behavior |
We welcome contributions of all kinds — bug reports, feature requests, docs improvements, and code. See our contributing guide to get started.
Website · Docs · Samples · Discussions
All projects are Apache 2.0 licensed.