Amazon Kindle
Read on Kindle.
For your Kindle or the Kindle app.
An illustrated guide to AI agents
AI agents and the systems around them, explained with a peanut.
Aishwarya Ashok (& agentic teammates)

12 chapters · 119 concepts
Think of this as an illustrated, dictionary-style handbook about AI agents. Each entry pairs a Peanut drawing with a short explanation of a term and what it means in practice.
The chapters cover models and agents, the tools and information they use, and what it takes to build and run them. Read a chapter for the bigger picture, browse a topic, or look up a term when you need it.
For anyone learning about AI agents, building with them, or working with a team that does. You don’t need a coding background to read it.
A worked example follows one replacement order through four stages, showing how ideas from across the book fit together. A closing system map and checklist help you connect the concepts and review an agent’s workflow.
A place to begin
These terms describe different parts of a system. A language model can sit behind a chatbot, an agent, or both.
The model generates a response from the input it receives. That response might be text or a proposed tool call. Naming a tool does not make the action happen; software has to run it.
A chatbot gives you a conversation: you send a message and get a reply. Behind that interface there may be a model, a predefined workflow, or an agent. The chat window alone does not tell you how the work is done.
An agent system uses a model to choose actions toward a goal. The surrounding software runs permitted actions and returns the results, giving the model information for its next decision.

Peanut represents the whole agent system here, choosing a repair, using a tool, and checking the result.
Give Peanut a wobbly chair. The model can suggest a repair; a chat interface lets you discuss it. An agent system can use tools, check the result, and decide whether another step is needed. The application sets the permissions and any required approvals.
Read the complete AI agent entry →Inside the complete book
The sample below spans six concepts. The complete edition contains all 12 chapters and a connected worked example.
A few pages to start with
Six complete entries, with illustrations, explanations, and notes on what Peanut represents in each drawing.
Read the sample from the beginning →Before you choose
Yes. The book explains the vocabulary, the parts of an agent system, and the decisions behind its behavior. You can follow the illustrations and explanations without writing code.
It is an illustrated, dictionary-style reference. Read the chapters in order or use the concept index to look up a term. A connected example follows one replacement order through four stages. The book does not include runnable code or a programming project.
It is written for product managers, designers, founders, students, and engineers getting familiar with agents. It helps you understand conversations about tools, memory, retrieval, approvals, evaluation, cost, and the workflow an agent serves.
Peanut makes a technical subject approachable. This handbook is written for readers learning about AI agents in product and engineering work; it is not a children’s storybook.
The free sample contains six complete entries: language models, AI agents, tool use, RAG, guardrails, and agent evaluations. Each includes the drawing, explanation, and a note on what Peanut represents.
Explore the free sample →Choose Kindle if you prefer reading in your Kindle library, on a Kindle device or in the Kindle app. The direct download includes a PDF that preserves the designed pages and a reflowable EPUB with adjustable text. Both direct formats have linked contents and a concept index.
Choose your edition →