Skip to content
← Tools / Skills
ToolAWS/10 trustMetrics-checked
Strands Agents logo

Strands Agents

AWS's open-source, model-driven agent SDK.

Visit official site →

About

Strands Agents is AWS's open-source SDK for building AI agents with a deliberately simple loop: you provide a model, a system prompt, and a list of tools, and the model plans, selects tools, and orchestrates steps itself instead of following hard-coded workflows.

It is Apache-2.0, ships SDKs for Python and TypeScript with 40+ pre-built tools, supports Bedrock, Anthropic, OpenAI, Gemini, and Ollama, and has native MCP support plus multi-agent primitives for handoffs and swarms. Multiple teams inside AWS already run it in production.

Limitation: it is newer than the established frameworks, so community examples are thinner, and the natural deployment path leans on Bedrock AgentCore.

What it can do

  • Model-driven agent loop
  • 40+ pre-built tools
  • Provider-agnostic: Bedrock, OpenAI, Anthropic, Gemini, Ollama
  • Native MCP support
  • Multi-agent handoffs and swarms
  • Session state and memory
  • Python and TypeScript SDKs
  • Deploys to Bedrock AgentCore Runtime

Setup

  1. Install the SDK: pip install strands-agents (Python 3.10+).
  2. Configure model credentials — Bedrock by default, or Anthropic, OpenAI, Gemini, Ollama via LiteLLM.
  3. Build an agent with Agent(tools=[...]) and pass MCP servers through MCPClient.
  4. Test locally, then deploy to Amazon Bedrock AgentCore Runtime for production hosting.

Prerequisites

Python 3.10+; model credentials (AWS Bedrock or another provider).

Use cases

agentsautomationcode

Works with

Generic APIAny agent (.md)

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: ~7.6k GitHub stars reported (Sept 2026); used in production by AWS teams (Amazon Q Developer, Glue reported). Manual review: skimmed AWS docs and SDK guides; not hands-on tested. License: Apache-2.0..

★ 7,600 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
7.0/10
Maintenance · recency, releases, issue load
8.0/10
Openness · license + docs
9.0/10
Manual review · sanity read + red-flag check
6.0/10
  • · ~7.6k GitHub stars reported (Sept 2026)
  • · used in production by AWS teams (Amazon Q Developer, Glue reported)
  • · manual review: skimmed AWS docs and SDK guides; not hands-on tested

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.