Skip to content

Strands harness quickstart

Strands harness is a state-of-the-art, fully assembled agent harness. You can get started in a few different ways:

Set up with your coding agent

Section titled “Set up with your coding agent”
Using a coding agent? Copy this prompt into Codex, Claude Code, Kiro, or any coding assistant and it will walk you through this page, ask which path, language, and model provider you want, and offer to set up the Strands MCP server.

Build an agent with the CLI

Section titled “Build an agent with the CLI”

The strands CLI gets you chatting with a harness agent in your terminal, with no code required.

Install the CLI

Section titled “Install the CLI”
Terminal window
npm install -g @strands-agents/cli

Create your agent

Section titled “Create your agent”

Run strands to open setup, then choose Quickstart. The only thing you set up is a model provider:

  1. Pick a provider. Choose Amazon Bedrock, Anthropic, OpenAI, Google Gemini, Ollama, or LiteLLM. The CLI detects credentials you already have, such as an AWS profile or an OPENAI_API_KEY in your environment. If a provider needs a key, paste it in. The CLI keeps a pasted key for the current session only, so add it to your shell profile to reuse it.
  2. Pick a model. Search the provider’s models, or keep the default. You can also set the reasoning effort and turn on web search here.
  3. Choose Save and Launch. Your agent opens in a chat.

Everything else comes from the default harness: file and shell tools, prompt caching, context management, long-term memory, and more. See the agent you get.

To change more than the model, choose Customize instead. It walks you through the agent’s name and instructions, tools, plugins, context and memory, and tool permissions. Run /setup in a chat to reopen setup later.

You can also start from an existing agent file, which skips setup, or override the saved configuration for a single run with flags:

Terminal window
strands --agent ./agent.ts                     # start from an exported agent
strands --model anthropic/claude-sonnet-5      # use a different model for this run

Export it to code

Section titled “Export it to code”

When you’re ready to embed the agent in an application, run /export in a chat. The command writes a Python or TypeScript project with your choices set directly on create_harness(...)createHarness(...). The project exports a ready-to-import agent, so the CLI is a fast on-ramp to the library below: build interactively, then drop into code.

Use it as a library

Section titled “Use it as a library”

Prefer to write the code yourself? Install the Python or TypeScript package and run your first agent in a few lines. The source lives on GitHub:

harness-py — the Python package.

Install

Section titled “Install”

Requires Python 3.10 or newer.

Terminal window
pip install strands-harness

Run your first agent

Section titled “Run your first agent”

Strands harness runs on the model of your choice and supports model providers across Amazon Bedrock, Anthropic, OpenAI, and Google, plus Ollama for running locally. Amazon Bedrock is the default; pass model="provider/name" to pick another. See choose a model for the full list.

Even a one-line call has a shell, file tools, and web access working out of the box. Strands harness can search the web, compare what it finds, and save the results to a file.

Bedrock is the default, using Claude Opus 5 in the region your AWS configuration selects.

from strands_harness import create_harness


agent = create_harness()
agent("Research the three most common strategies for versioning a REST API, compare their tradeoffs, and write a recommendation to api-versioning.md")

Give Strands harness AWS credentials with permission to invoke the model, using one of:

  • Bedrock API key: set AWS_BEARER_TOKEN_BEDROCK to a Bedrock API key. Quickest for local development.
  • AWS credentials: aws configure, or AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and optionally AWS_SESSION_TOKEN as environment variables.
  • IAM roles: on AWS services like EC2, ECS, or Lambda.

Enable access to the models you use in the Amazon Bedrock console; see the AWS documentation.

Keep a conversation across runs

Section titled “Keep a conversation across runs”

Sessions are on by default: Strands harness persists each conversation to disk under ./.agent/sessions with a generated id. Choose the id yourself and a later run rehydrates the same conversation:

from strands_harness import create_harness


agent = create_harness(session={"id": "api-design"})
agent("Which of those would you pick for an API with external customers, and why?")

See persist sessions for how session storage works and how it differs from long-term memory. The CLI persists sessions the same way — resume one with strands --session-id api-design.

The agent you get

Section titled “The agent you get”

However you start it, Strands harness runs on the model of your choice with reasoning on, and it follows a tuned system prompt that tells it to explore before changing things, confirm before anything irreversible, and verify before calling a task done. It also has prompt caching, automatic context management, long-term memory, a generalist subagent, and a todos task tracker enabled by default. See what the default harness does for the full set.

Next steps

Section titled “Next steps”