Strands Agents

Tracked

Open-source SDK for production AI agents in Python and TypeScript — build an agent harness and control it end-to-end, any model, any cloud.

Author strands-agents Open Sourced 2025-05-14 Last Commit Unknown

Overview

Strands Agents is an open-source SDK for building production AI agents in Python and TypeScript. It lets you build an agent harness and control it end-to-end, working with any model (Anthropic, OpenAI, Bedrock) and any cloud. It supports MCP tool integration, multi-agent systems, and autonomous agent workflows.

Key Features

  • Agent harness SDK with end-to-end control in Python and TypeScript.
  • Model-agnostic: Anthropic, OpenAI, AWS Bedrock, and more.
  • First-class MCP tool support.
  • Multi-agent systems and autonomous agent patterns.

Use Cases

  • Build production agent harnesses with full lifecycle control.
  • Deploy agents across clouds and model providers without lock-in.
  • Compose multi-agent systems with MCP tooling.

Technical Details

  • Harness-centric design: the SDK exposes the full agent loop — tools, hooks, and session state — rather than hiding it behind opaque abstractions.
  • Model-agnostic driver layer works with Anthropic, OpenAI, and AWS Bedrock behind one interface.
  • Tools follow the MCP standard, so existing MCP servers plug in directly.
  • Dual-language: Python and TypeScript SDKs share the same concepts for cross-stack teams.