LangGraph and Strands Agents: Core Concepts, Patterns, and Tradeoffs
FORMAT: WorkshopLEVEL: AdvancedLANGUAGE: Spanish
The agent framework landscape has consolidated fast. Among the options available today, two stand out for their architectural depth and production relevance: LangGraph, with its graph-based, developer-controlled orchestration model; and Strands Agents, with its model-driven, minimal-scaffolding approach. They are not interchangeable — they represent different answers to the same fundamental question of how much control a developer should retain over agent behavior.
This workshop is a structured, hands-on exploration of both. Participants will work through a shared codebase covering real scenarios: tool calling, state management, multi-turn reasoning, and multi-agent coordination. Each scenario is implemented in both frameworks side by side, making the architectural differences concrete and comparable rather than theoretical.
The session is organized around three axes: core concepts (how each framework models state, tools, memory, and control flow), practical patterns (what each one makes easy, what it makes hard), and honest tradeoffs (where each framework earns its complexity and where it doesn't). No polished demos — just real code, real outputs, and an open discussion of what each approach costs and buys you.
By the end, participants will have a working mental model for both frameworks, hands-on exposure to their key patterns, and enough grounded perspective to make an informed architectural decision on their next agentic project.
Suitable for: Python developers with some familiarity with LLMs or agent concepts. Production experience is a plus but not required — the workshop is designed to reward depth of engagement, not prior framework knowledge.
