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.
What is 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.
Overview of harness-sdk
What it is
Strands Agents is an open-source SDK for building and running AI agents in Python and TypeScript. It provides a model-driven approach to constructing agent harnesses that run in your own process with no hosted control plane. The project includes a fully assembled harness via create_harness() (Python) or createHarness() (TypeScript), as well as lower-level SDKs for custom agent loops. It supports any model, any cloud, and covers lifecycle controls, tools, structured output, MCP, multi-agent patterns, memory, sessions, streaming, guardrails, tracing, and evals.
Key features
- Model agnostic: First-class support for Amazon Bedrock, Anthropic, OpenAI, Gemini, plus many more providers and custom ones.
- Lifecycle controls: Turn limits, token budgets, cancellation, stop reasons, and hooks to intercept any step for logging, validation, or redirection.
- Built-in observability: The agent loop traces every decision by default.
- Guardrails and steering: Guardrails catch mistakes before they run; steering handlers let agents correct themselves instead of failing silently.
- Multi-agent patterns, MCP, streaming, and structured output are all built in.
Use cases
Use Strands when you would otherwise write your own agent loop. It is suitable for production AI agents that need end-to-end control, model portability, and built-in observability. Developers can start with the batteries-included harness for quick prototyping and then drop down to the SDKs to own the agent loop, tools, model providers, and memory.
Getting started
Python harness:pip install strands-harnessTypeScript harness:npm install @strands-agents/harnessFor the lower-level SDK, see the repository README.
License and pricing
Licensed under the Apache License 2.0. The project is open-source.
Based on the project's public README and repository metadata.