Strands Harness SDK

New to Strands?

Start with Strands harness for a fully assembled, state-of-the-art agent harness. It’s built on the Strands Harness SDK with opinionated, benchmarked defaults for the system prompt, tools, memory, sessions, and context management, and every default is overridable.

Use the Strands Harness SDK when you want to build an agent harness from the ground up, piece by piece.

An agent harness is the software around a model that turns it into an agent. The Strands Harness SDK is for building one yourself, end to end.

It runs the loop and gives you the pieces you attach to it: tools, memory, sessions, plugins, and interventions. The same code takes you from prototype to production.

A running agent

The smallest agent is a model with the loop around it. Create one and invoke it:

Python

from strands import Agent

agent = Agent()
agent("Explain the agent loop in one sentence.")

TypeScript

import { Agent } from '@strands-agents/sdk'

const agent = new Agent()
const result = await agent.invoke('Explain the agent loop in one sentence.')
console.log(result.lastMessage)

From here you add the tools, memory, and model provider your application needs, then take it to production.

Strands harness or the Strands Harness SDK?

Strands gives you two starting points, and you can move between them without a rewrite.

If you want to…	Start with	Why
Get a complete agent harness with tested defaults	Strands harness	The decisions are already made: system prompt, tools, memory, sessions, context management, and a production loop come pre-wired and benchmarked. You start from a working agent.
Build the agent harness yourself, piece by piece	Strands Harness SDK quickstart	Nothing is pre-decided. You get the primitives and the loop, and you choose every component and how they fit.
Start on Strands harness and drop down for more control later	Compose with the Strands Harness SDK	Strands harness is built on the Strands Harness SDK, so any piece is yours to swap out without a rewrite.

Weighing Strands against other frameworks or a hand-written loop instead? See choosing an agent foundation.

QuickstartWalk through installing the Strands Harness SDK, choosing a model provider, and running your first agent in Python or TypeScript.

Build

Add toolsGive the agent capabilities with community, custom, and MCP tools.

Give it long-term memoryPersist and recall knowledge across sessions.

Return structured outputGet typed, schema-validated results back from the agent.

Coordinate multiple agentsCompose agents as tools, graphs, and swarms.

Run

Deploy to productionShip to Lambda, Fargate, EKS, Bedrock AgentCore, and more.

Observe your agentTrace runs, read metrics, and debug behavior.

Secure for productionAdd guardrails, redact PII, and keep message history trusted.

Where to go next

Building a feature? Each build guide above takes one task end to end. When you need to look up how a piece works, the Components section is the reference underneath the guides: the agent loop, models, state, sessions, hooks, plugins, interventions, and realtime.