Strands Agents lets you build AI research assistants in 30 lines of code

Strands Agents, an open-source Python framework from AWS, slashes the complexity of building intelligent AI agents. Instead of wrestling with multiple API calls, conversation state, and custom orchestration, developers declare agents with just a prompt and a tools list — the framework’s LLM handles reasoning, planning, and tool selection automatically. A complete research assistant can be built in 30 lines of code using the @tool decorator to integrate external APIs. The framework supports single agents, multi-agent networks, and hierarchical systems, all while running identically in local dev and production, with real-time streaming responses for interactive applications.

Strands Agents integrates natively with AWS services — Amazon Bedrock provides access to foundation models like Anthropic’s Claude Sonnet, and the framework works with AWS Lambda for serverless deployment. The Kiro IDE bundles over 50 curated “powers” (pre-built MCP servers, steering files, and API patterns) for one-click installation, covering design, security, and observability. This combination lets developers skip infrastructure wrangling and focus on the agent’s actual intelligence. Already used internally by AWS teams in Amazon Q and AWS Glue, Strands Agents is Apache-2.0 licensed with active community contributions, making it a production-ready choice for building anything from research assistants to automated analysis pipelines.