🧠 Build AI Agents: Think-Plan-Act with Llama-4 Reasoning

This n8n workflow template implements a sophisticated 'Think-Plan-Act' architecture for building AI agents using Llama-4 via OpenRouter or similar providers. It enables automations to analyze tasks, generate step-by-step plans with subgoals and tools, and execute actions intelligently, returning structured JSON for seamless integration.

Key components include a modular agent loop with THINK mode for reasoning (goals, subgoals, tools, assumptions) and ACT mode for step-by-step execution. Built-in parsers ensure reliable JSON outputs with natural language fallbacks. A demo workflow (e.g., 'get the weather') demonstrates quick setup.

Benefits: Enhances any n8n flow with tool-aware reasoning, composability, and customization. Saves hours of development by providing prompts, schemas, and logic. Ideal for complex automations like data analysis, API orchestration, or multi-step processes.

Use cases: Integrate into CRM, e-commerce, or research workflows for smarter decision-making. Customize goals, tools, and prompts easily—no advanced coding needed. Perfect for developers, analysts, and teams building agentic AI.