Agents 101 guide
FreeIntroductory guide to LLM agents and multi-agent systems
FreeFree tier
About Agents 101 guide
An introductory guide to LLM agents from the awesome-generative-ai-guide repository. It defines an LLM agent as a reasoning core (the model) plus a harness (tools, memory, and a loop for iterative task completion). The guide breaks down the harness into four components: agent core (decision loop), memory (short-term and long-term with context engineering), tools (e.g., web search, code execution, RAG, MCP), and planning (task decomposition and reflection). It also covers multi-agent systems for division of labor across specialized agents. Includes a practical vacation planner example and references to NVIDIA and survey sources.
Key Features
Defines LLM agents as model plus harness
Breakdown of harness: agent core, memory, tools, planning
Covers multi-agent systems and division of labor
Includes practical vacation planning example
References NVIDIA and survey sources
Discusses context engineering and Model Context Protocol (MCP)
Pros & Cons
Pros
- Comprehensive breakdown of agent components
- Includes multi-agent systems and coordination
- Practical example illustrates concepts clearly
- Well-referenced with industry sources
- Free and open source
Cons
- Guide is introductory and lacks implementation details or code examples
- Assumes familiarity with large language models
- Limited depth on advanced topics like reasoning models
Best For
Learning about LLM agent architectureUnderstanding components like memory, tools, and planningDesigning multi-agent collaborative systemsReference for building production-ready agentsEducational resource for AI engineers
FAQ
What is an LLM agent?
An LLM agent is a large language model given the ability to act, not just answer. It consists of a model (the reasoning core) plus a harness (tools, memory, and a loop that lets it try, observe, and try again until the task is done).
What are the key components of an agent harness?
The harness includes four parts: agent core (decision loop), memory (short-term and long-term), tools (actions like web search, code execution, RAG), and planning (task decomposition and reflection).
What is a multi-agent system?
A multi-agent system splits a problem across several specialized agents that collaborate on the principle of division of labor, improving efficiency and quality on complex tasks.