Strands Agents
Open source toolkit for building AI agent harnesses in Python and TypeScript
- Tool type
- Open source agent toolkit and pre-built harness
- Made for
- Developers building AI agents
- Runs on
- Python, Node.js and CLI environments
- Agents run in your own Python or Node.js process or the terminal; models run on Bedrock, Anthropic, OpenAI, Google, or locally via Ollama.
What it does
Strands is an open source toolkit for building AI agents in Python and TypeScript. You give it a task in plain language and choose a model: Amazon Bedrock by default, or Anthropic, OpenAI, Google, or Ollama. The agent then runs a loop that calls tools, including shell and file tools (read, write, edit), web access, and custom functions you write. Output is the agent's response plus side effects such as files written to disk.
Strands harness is a pre-assembled agent you can start with a single call, with defaults for tools, context management, session persistence, and memory. The Strands Harness SDK is the lower-level layer for assembling an agent piece by piece. Custom tools can be added with a decorator, MCP tools can be attached, and a terminal CLI can start a chat agent without code and export it to Python or TypeScript.
It suits developers who want a model-agnostic agent starting point they can customize. Setup needs Python 3.10+ or Node.js 20+, plus credentials for the chosen model provider. Model and compute costs are separate from the toolkit. Check the GitHub repositories for licensing terms before commercial use.
What it can do
- Runs an agent loop against a chosen model provider (Amazon Bedrock, Anthropic, OpenAI, Google, Ollama)
- Provides built-in shell and file tools (read, write, edit) and web access in the default harness
- Manages context windows and caches reused request parts
- Persists conversations across runs using a session id
- Lets developers add custom tools with a decorator and attach MCP tools
- Exports CLI-built agents to Python or TypeScript code
Works with
- Amazon Bedrock
- Anthropic
- OpenAI
- Google Gemini
- Ollama
- LiteLLM
- Model Context Protocol (MCP) servers
Example workflows
- Research and write a comparison file
- Input: a natural-language task such as researching three API versioning strategies. Steps: create the harness with create_harness(), then invoke the agent with the task. Output: the agent uses web and file tools and writes a markdown file such as api-versioning.md.
- Resume a conversation across runs
- Input: a session id such as 'api-design'. Steps: create the harness with that session id, then ask a follow-up question that depends on earlier context. Output: the session is stored under ./.agent/sessions and rehydrated for the follow-up answer.
Before you choose
- Bedrock is the default model provider, so model access and AWS credentials must be configured before the default agent runs.
- Strands harness is opinionated but overridable; the lower-level Strands Harness SDK is needed for a fully custom build.
- Review the GitHub repositories for license terms before commercial use; model provider and compute costs are billed separately.
Listed under
Get started
Start with the quickstart pages; each path requires a model provider and its credentials.
- Install the Python package with pip install strands-harness (Python 3.10+) or the TypeScript package with npm install @strands-agents/harness (Node.js 20+).
- Choose a model provider and set its credentials, such as AWS credentials or AWS_BEARER_TOKEN_BEDROCK for Bedrock, or ANTHROPIC_API_KEY or OPENAI_API_KEY for those providers.
- Run a first agent with create_harness() in Python or createHarness() in TypeScript, then invoke it with a task.
Plans and prices
The source code is available under Apache 2.0. Model-provider usage and compute are separate.
- Apache 2.0 license
Source: github.com/strands-agents/sdk-python, checked 28 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.
- Harness SDK | Strands Agents ↗strandsagents.com
- Python Quickstart | Strands Agents ↗strandsagents.com
- Strands harness | Strands Agents ↗strandsagents.com
- Strands harness quickstart | Strands Agents ↗strandsagents.com
- GitHub - strands-agents/harness-sdk: Build an agent harness and control it end-to-end. Open-source SDK for production AI agents in Python & TypeScript - any model, any cloud. · GitHub ↗github.com
Ratings and reviews
No ratings for Strands Agents yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- strandsagents.com: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Web search
- Confirms the official site
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.