Strands Agents (AWS)
Strands Agents is an open-source AWS project and toolkit for building production AI agents. Its model-driven approach connects a model, tools, and prompts so agents can plan, reason, call tools, collaborate with other agents, and move from local development to production.
At a Glance
- Developers and open-source communities building AI agents
- Enterprise engineering and platform teams moving agents from prototypes to production
- AWS customers using Amazon Bedrock, Lambda, EC2, Fargate, or AgentCore
- Organizations building assistants, research agents, software-development agents, and operational automation
- +2 more
AI Tools by Strands Agents (AWS)
(2)
Strands Decider
Open Source Agent Decision Model
Strands Agents
AWS Open Source Agent SDK
Discussions
No discussions yet
Be the first to start a discussion about Strands Agents (AWS)
Latest News
Introducing Strands Decider 2B, a small open-source decision model
Introducing Strands harness with frontier performance and 28% lower token cost
Framework-agnostic evaluation with Strands Evals
Strands Agents production SDK reaches 99.95% on ARC-AGI-3
Products & Services
Apache-2.0 open-source, model-driven SDK for building and running AI agents with prompts, models, tools, agentic loops, multi-agent patterns, MCP, A2A, memory, sessions, and customizable context management.
TypeScript/JavaScript SDK for type-safe agents in Node.js and the browser, with model providers, custom tools, plugins, and multi-agent orchestration.
A fully assembled, customizable agent harness with optimized defaults for prompts, tools, memory, sessions, and context management, available as a CLI or library.
Lower-level SDK for building an agent harness from the ground up, including the agent loop, tools and MCP, model providers, memory, sessions, and hooks.
Market Position
Strands differentiates itself from workflow-heavy agent frameworks through a model-driven loop in which the foundation model determines planning, tool selection, and iteration. It combines AWS-native production paths with an Apache-2.0, model-agnostic open-source implementation that can also use Anthropic, Meta/Llama, Ollama, OpenAI via LiteLLM, and custom providers; its positioning is therefore both cloud-integrated and portable rather than tied exclusively to one model vendor.
Leadership
Founders
Dr. Swami Sivasubramanian
Vice President of Data and AI at Amazon Web Services (AWS). Leading AWS's efforts in generative AI, databases, and analytics.
AWS Agentic AI Team
The internal engineering group at AWS responsible for developing agentic orchestration and observability tools.
Executive Team
Matt Garman
CEO, Amazon Web Services
Leading AWS strategy and operations; formerly Senior VP of AWS Sales, Marketing, and Support.
Dr. Swami Sivasubramanian
VP of Data and AI, AWS
Oversees the entire AI/ML stack at AWS, including Bedrock and SageMaker.
Board of Directors
Founding Story
The AWS team behind Amazon Q Developer began building agents in early 2023 and found that existing frameworks required months of prompt, parser, and orchestration tuning. As models gained native reasoning and tool-use capabilities, the team created Strands Agents to remove unnecessary orchestration complexity; internally, Q Developer teams were able to move from prototype to production in days or weeks rather than months.
Business Model
Revenue Model
Strands Agents is distributed as free, open-source Apache-2.0 software rather than as a separately priced hosted product. Users bring their model/provider and deployment environment; AWS integrations can run through services such as Amazon Bedrock and Bedrock AgentCore, whose usage is separately billed by AWS.
Target Markets
- Developers and open-source communities building AI agents
- Enterprise engineering and platform teams moving agents from prototypes to production
- AWS customers using Amazon Bedrock, Lambda, EC2, Fargate, or AgentCore
- Organizations building assistants, research agents, software-development agents, and operational automation
- Robotics and physical-AI researchers and developers
- Teams needing cross-framework evaluation and agent safety testing
- Automating adaptive business workflows and classification, scoring, and routing
- AI assistants grounded in private data through retrieval tools
- Autonomous multi-step web research and synthesis
- Software-development assistance and operational troubleshooting
- Multi-agent collaboration and orchestration
- Agent evaluation, simulation, guardrails, and red teaming
- Amazon Q Developer
- AWS Glue
- VPC Reachability Analyzer
- Landchecker