Introduction

Agent engineering is the discipline of designing AI agents that work for you — not by writing code, but by writing instructions, connecting the right tools, and structuring information so that an AI can act on it effectively.

What is an agent?

An AI agent is a system built on a large language model (LLM) that can not only think and respond, but also take actions. Unlike a simple chatbot that just answers questions, an agent can use tools, access files, connect to your systems, and get things done on your behalf.

What makes an agent powerful is not magic — it is a simple loop, good instructions, and the right tools. This guide breaks that down.

What is agent engineering?

Agent engineering is not software engineering.

You should not need to write code. But you do need to understand how agents work well enough to design them effectively. That means writing clear instructions, choosing what tools and data the agent needs, and structuring how it interacts with the world.

The best agent engineers are not necessarily software engineers. They are domain experts — people who deeply understand the work the agent needs to do and can translate that understanding into well-designed agent systems.

Your agent platform or tech team should handle the technical details. You handle the design — because nobody understands the work better than you.

What this guide covers

This guide covers the fundamental concepts of agent engineering. These concepts are meant to work with any agent platform — this is not a step-by-step tutorial for a specific one. Understanding them will make you effective regardless of which platform you choose.

However, this is not a comprehensive guide to everything that agents can be. Rather, it is an opinionated set of the smallest number of concepts that enable you to build extremely powerful agents.

The guide introduces five core concepts, each building on the last:

1. Agents — The core loop: how an agent thinks, acts, and responds. How to write the instructions that define what your agent is and does.

2. Tools — How agents interact with the world: accessing files, running commands, and connecting to external systems.

3. Skills & Volumes — How to customize your agent with on-demand instructions and connect it to the right data.

4. Bots — How to make your agent available where you already communicate: Telegram, Slack, WhatsApp, and more.

5. Proactive Agents — How to make your agent act on its own through schedules and heartbeats, so it works for you even when you are not asking.

Each concept is explained simply, with concrete examples. No unnecessary technical details — just what you need to understand to design agents that work.

How to take this into practice

Once you understand the concepts, you will want a platform to build on. Here are a few options:

• CloudNova — Built from the ground up around the concepts in this guide. Full disclosure: we are the creators of CloudNova. The screenshot examples throughout this guide are from CloudNova.

• OpenClaw — Arguably the most popular agent platform today, with a large community and extensive integrations.

These are just a few examples — the landscape is growing fast and there are many others. That is exactly why understanding the underlying concepts matters more than mastering any single platform. Platforms will come and go. The fundamentals of designing great agents will not.

Ready? Let’s get started →