來源AWS Machine Learning Blog•較早收集於 2m
使用 Strands 輕鬆構建智慧研究助理
別再為複雜的代理編排苦惱；看看 Strands 如何簡化智慧研究助理的開發。
30 秒速覽
有什麼變化
降低 AI 代理多重 API 呼叫編排的複雜度
為什麼重要
此工具顯著降低了構建專業 AI 研究代理的門檻，有望加速特定領域 AI 工具的部署。
下一步行動
查看 Strands 文件，評估它是否能取代您目前研究代理專案中自建的編排層。
誰應關注：Developers & AI Engineers
關鍵要點
- •降低 AI 代理多重 API 呼叫編排的複雜度
- •自動化研究導向應用程式的對話狀態管理
- •讓開發者無需高階機器學習學位即可構建推理代理
深度解析
背景與延伸：來自公開資料，非原文內容。引用 14 個來源。
增強重點摘要
- •Strands Agents is an open-source SDK from AWS, released in May 2025, that employs a model-driven approach, allowing Large Language Models (LLMs) to autonomously plan and use tools based on a simple prompt and tool list, significantly reducing the need for complex hardcoding.
- •The framework is model-agnostic, supporting various LLM providers including Amazon Bedrock, Anthropic, OpenAI, Ollama, and Meta (via LiteLLM), and offers flexible deployment options from local development to AWS services like Lambda and Bedrock AgentCore.
- •Strands facilitates multi-agent collaboration patterns such as Agents as Tools, Swarms Agents, and Agent Graphs, and integrates natively with AWS services, being actively used internally by AWS teams for production services like Amazon Q and AWS Glue.
- •AWS launched Strands Labs in March 2026, a GitHub organization for experimental agent-based AI projects, including robotics integration (Strands Robots, Robots Sim) and AI Functions for code generation, indicating future directions beyond traditional research assistants.
技術深入
- Model-Driven Core: Strands leverages the reasoning capabilities of modern LLMs as the central intelligence to autonomously handle planning, decision-making, and tool usage, rather than relying on hardcoded workflows.
- Key Components: Building an agent with Strands primarily requires a language model, a system prompt defining the agent's role, and a list of tools it can utilize.
- Tool Integration: External functions and APIs can be easily integrated as tools using a Python
@tooldecorator. It supports the Model Context Protocol (MCP) for accessing thousands of external tools and the Agent-to-Agent (A2A) protocol for inter-agent communication. - Agentic Loop: The SDK implements a lightweight and extensible agent loop that invokes the LLM with the prompt and agent context, along with descriptions of available tools, enabling the LLM to dynamically choose and execute actions.
- Deployment Flexibility: Strands agents can run locally for development and be deployed to production environments as a monolith or microservices, leveraging AWS services like Lambda, ECS/Fargate, and Bedrock AgentCore for serverless execution and managed runtime.
- Observability and Evaluation: The framework integrates with SageMaker Serverless MLflow for agent tracing and offers Strands Evals, including ToolSimulator, for systematic evaluation and safe testing of agents that interact with external tools.
- Multi-Agent Architectures: Strands supports advanced multi-agent collaboration patterns, such as Agents as Tools, Swarms Agents, Agent Graphs, and Agent Workflows, allowing for the composition of specialized sub-agents to tackle complex, multi-faceted tasks.
前景展望基於引用來源的 AI 分析
Strands will accelerate the adoption of AI agents in enterprise environments.
Its simplified, model-driven approach and native AWS integration lower the barrier to entry for developers to build and deploy complex AI agents for various business use cases.
The framework will expand into more specialized domains, particularly robotics and advanced code generation.
The introduction of Strands Labs with projects like Robots, Robots Sim, and AI Functions indicates a strategic move towards integrating AI agents with physical hardware and automating software development.
Strands will foster a more modular and collaborative approach to AI system design.
Its support for multi-agent collaboration patterns and the Agent-to-Agent (A2A) protocol encourages the development of specialized, interoperable agents rather than monolithic AI systems.
時間線
2025-05
AWS releases Strands Agents, an open-source SDK.
2025-07
AWS publishes a technical deep dive into Strands Agents SDK and its architectures.
2025-07
AWS demonstrates building a drug discovery research assistant using Strands Agents and Amazon Bedrock.
2025-07
AWS demonstrates building dynamic web research agents with Strands Agents SDK and Tavily.
2025-11
AWS explores multi-agent collaboration patterns with Strands Agents and Amazon Nova.
2026-03
AWS introduces Strands Labs for experimental AI agent projects, including robotics and AI Functions.
- 2025-05AWS releases Strands Agents, an open-source SDK.
- 2025-07AWS publishes a technical deep dive into Strands Agents SDK and its architectures.
- 2025-07AWS demonstrates building a drug discovery research assistant using Strands Agents and Amazon Bedrock.
- 2025-07AWS demonstrates building dynamic web research agents with Strands Agents SDK and Tavily.
- 2025-11AWS explores multi-agent collaboration patterns with Strands Agents and Amazon Nova.
- 2026-03AWS introduces Strands Labs for experimental AI agent projects, including robotics and AI Functions.
來源 (14)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
1dev.tovertexaisearch.cloud.google.com2amazon.comvertexaisearch.cloud.google.com3medium.comvertexaisearch.cloud.google.com4amazon.comvertexaisearch.cloud.google.com5amazon.comvertexaisearch.cloud.google.com6amazon.comvertexaisearch.cloud.google.com7amazon.comvertexaisearch.cloud.google.com8cloudthat.comvertexaisearch.cloud.google.com9amazon.comvertexaisearch.cloud.google.com10infoq.comvertexaisearch.cloud.google.com11missioncloud.comvertexaisearch.cloud.google.com12medium.comvertexaisearch.cloud.google.com13aws.comvertexaisearch.cloud.google.com14amazon.comvertexaisearch.cloud.google.com
AI 週報
閱讀本週精選 AI 大事摘要 →
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： AWS Machine Learning Blog ↗
每週電子報
每週一封，可隨時退訂。