AI agent
An AI agent is a computational system that uses artificial intelligence to interpret information, select actions, and pursue goals through interaction with an environment.
Introduction to Generative AI, Second Edition
An autonomous agent is a software system that independently perceives an environment, selects actions, and executes tasks with limited direct human intervention.
- How it is used
- Positions agents as a way to delegate multistep tasks across tools while retaining human oversight for consequential decisions.
- How deep it goes
- Defines agents through sensing and action, then explains tool chaining and oversight without implementation-level instructions.
- Examples
- Uses a thermostat, a weather-and-calendar outfit request, and a sales-data-to-web-search task to illustrate increasing complexity.
- What you can do after
- Evaluate/understand it by distinguishing autonomous action from fixed commands and identifying when permissions, approval, and monitoring matter.
Read how Introduction to Generative AI, Second Edition explains AI agent →
Grokking AI Applications
Artificial intelligence whose systems can operate autonomously, make decisions, and perform tasks with limited human intervention.
- How it is used
- Extends LLM applications when they must fetch live data, trigger actions, or make multistep decisions.
- How deep it goes
- Explains the reasoning-and-tool loop, then builds calculator and travel-agent flows without model-internals detours.
- Examples
- Uses a sleep-comfort controller, calculator, HistoryTravelBuddy itinerary planner, and recipe-search chef agent.
- What you can do after
- Apply it via Langflow: configure an Agent component, toolify components, and connect tools for autonomous task completion.
The Art of AI Product Development
An AI agent is an artificial intelligence system that perceives relevant conditions and takes actions toward goals.
- How it is used
- Positions agents as a way to automate multistep workflows that otherwise require people to plan, use tools, and coordinate results.
- How deep it goes
- Explains the environment interaction cycle and compound architecture, then gives incremental design guidance without deriving model internals.
- Examples
- Uses a product-roadmap workflow involving interviews, usage data, and stakeholder feedback, alongside an agent–environment interaction diagram.
- What you can do after
- Evaluate/understand it by scoping a small agent, testing quality and latency, and deciding when actions require human review.
Read how The Art of AI Product Development explains AI agent →
AI Agents in Action, Second Edition
Software that perceives an environment, selects actions, and acts toward a goal using resources provided by a large language model.
- How it is used
- Turns an LLM-powered application from answering requests toward planning and carrying out multistep goals.
- How deep it goes
- Develops a sense–plan–act–learn loop and five functional layers, then progresses through a minimal agent, typed output, tools, and tracing.
- Examples
- Uses flight booking and a research-planning exercise with numbered tasks, typed output, a retrieval tool, and traces.
- What you can do after
- Build it from scratch as a minimal research-planning agent, then add structured output, tool use, and tracing.
Read how AI Agents in Action, Second Edition explains AI agent →
Inside AI
An autonomous system is a system that performs functions and makes operational decisions with limited direct human control.
- How it is used
- Uses smart agents to frame distributed, adaptive responses to changing problems, especially fraud and security monitoring.
- How deep it goes
- Defines agent capabilities and contrasts them with state-space search, but leaves full technical treatment outside its overview.
- Examples
- Illustrates agents with collaborating financial-security specialists, continuously updated fraud profiles, and a proprietary platform.
- What you can do after
- Evaluate/understand where cooperating, adaptive agents might help and why autonomy still requires attention to operational risk.
Domain-Specific Small Language Models
An autonomous software entity that can invoke models or tools as part of a larger system.
- How it is used
- Deploys language models as autonomous, tool-using systems for multistep tasks and domain-specific automation.
- How deep it goes
- Explains reasoning, acting, interaction, memory, planning, and tool chaining, while deferring advanced agent concepts.
- Examples
- Demonstrates flight booking, custom Python tools, CodeAgent workflows, and adaptive hybrid retrieval over local documents.
- What you can do after
- Build it from scratch with smolagents, Python-based tools, small language models, and chained local or web retrieval.
Read how Domain-Specific Small Language Models explains AI agent →
AI Agents and Applications
An AI agent is a software system that works with a large language model to perform multi-step tasks by selecting actions, invoking tools, and adapting to intermediate results.
- How it is used
- Coordinates open-ended, multi-step requests by selecting tools and adapting actions to intermediate results.
- How deep it goes
- Develops tool calling from a single-tool implementation through multi-tool agents, manual protocol handling, prebuilt components, and debugging.
- Examples
- Uses holiday-package planning and a Cornwall travel assistant with destination retrieval and weather results.
- What you can do after
- Build it from scratch with tool calls, then apply it via LangGraph components and debug execution traces.
AI Agents in Action
An agent system is a software system in which one or more AI agents perform tasks through reasoning, interaction, and access to tools or data.
- How it is used
- Frames agents as practical software interfaces that automate LLM interaction, task execution, information access, and collaboration.
- How deep it goes
- Explains interaction types, components, planning, iteration, evaluation, and feedback, while deferring detailed implementation to later techniques and frameworks.
- Examples
- Uses ChatGPT, DALL-E 3, AutoGPT, CrewAI joke agents, sequential crews, planning exercises, and feedback-loop experiments.
- What you can do after
- Build it from scratch by defining goals, profiles, tasks, tools, plans, execution modes, evaluation, and feedback loops.
Build an AI Agent (From Scratch)
An AI agent is a software system that uses artificial intelligence to perceive information, perform tasks, and make or execute decisions on a user's or organization's behalf.
- How it is used
- Directs multistep tasks whose next actions depend on information discovered during execution.
- How deep it goes
- Explains the LLM–tool–loop mechanism and develops a ReAct implementation with state, error handling, and evaluation.
- Examples
- Uses adaptive research, a search-and-calculation question, a Python calculator call, and GAIA problems.
- What you can do after
- Build it from scratch as a tool-using ReAct agent, then evaluate and debug its multistep runs.
Read how Build an AI Agent (From Scratch) explains AI agent →
Knowledge Graphs and LLMs in Action
An agent is a computational system that can use information and perform actions to accomplish tasks.
- How it is used
- Deploys LLMs with prompts, memory, retrieval, and tools to answer complex questions and perform adaptive enterprise workflows.
- How deep it goes
- Explains agent components, grounding patterns, shared-state pipelines, routing, retries, and implementation examples, while deferring exhaustive production engineering.
- Examples
- Uses conversational chat, knowledge-graph question answering, vector and graph retrieval, text-to-Cypher generation, query retries, and result summarization.
- What you can do after
- Build it from scratch as a conversational Python agent or implement modular retrieval and graph-query workflows with LangGraph.
Read how Knowledge Graphs and LLMs in Action explains AI agent →
Data Analysis with LLMs
An agent is a software pattern in which a language model dynamically decomposes a task, selects actions or tools, and plans subsequent steps from their results.
- How it is used
- Uses agents for data-analysis questions whose required data sources and sequence of steps cannot be fixed beforehand.
- How deep it goes
- Explains the tool-selection loop and walks through LangChain setup and tool traces, without examining model internals.
- Examples
- Traces a video-game sales query followed by web search, and a revenue query followed by currency conversion.
- What you can do after
- Apply it via LangChain to configure and invoke a database-and-web agent for a user question.
Authorization in Action
A software agent is a software system that acts autonomously or semi-autonomously on behalf of a person, organization, or another system.
- How it is used
- Governs agents that analyze data, retrieve documents, and automate enterprise workflows through external tools.
- How deep it goes
- Explains runtime planning, tool invocation, policy enforcement, and constraint-aware planning without detailing model internals.
- Examples
- Uses document search, customer support retrieval, procurement requests, and blocked access to credential files.
- What you can do after
- Operate/configure it with runtime policy checks that constrain tool calls and inform later planning.
LLMs in Production
A software system that uses artificial intelligence to perceive information, make decisions, and perform actions toward goals.
- How it is used
- Uses agents to extend LLM applications from answering requests to completing multistep tasks.
- How deep it goes
- Explains LLM, memory, and tools with memory trade-offs and LangChain examples, while noting that reliable behavior remains difficult.
- Examples
- Shows coding and search agents, Python and CSV tool calls, email replies, and résumé extraction.
- What you can do after
- Apply it via LangChain to assemble a tool-using agent, while accounting for unreliable output and execution risks.
Investing for Programmers
An AI agent is an artificial intelligence system that independently plans and performs multiple steps toward a specified task.
- How it is used
- Automates structured financial research by coordinating data collection, analysis, evaluation, and reporting.
- How deep it goes
- Explains planning, tool use, reflection, frameworks, and implementation choices, but defers production reliability and comprehensive validation.
- Examples
- Demonstrates asset research, stock-price retrieval, sentiment analysis, Microsoft news analysis, n8n workflows, and Notion exports.
- What you can do after
- Build structured research workflows with prompts, external tools, n8n, or LangChain, then validate their outputs cautiously.
Build AI-Enhanced Web Apps
An AI agent is a software agent that uses artificial intelligence to perceive inputs, make decisions, and produce actions toward a task.
- How it is used
- Powers task-oriented web experiences that decide when to use external tools rather than only generate replies.
- How deep it goes
- Works through LangChain.js agent components and assembly, while treating model selection and system prompts as previously covered.
- Examples
- Uses database sales queries, web search, and a LangChain.js Wikipedia tool to illustrate agent actions.
- What you can do after
- Build it from scratch with LangChain.js by configuring a model, system prompt, and tools, then integrating the agent into a chain.
Software Testing with Generative AI
An AI agent is a software system that uses artificial intelligence to pursue goals by interpreting information and taking actions.
- How it is used
- Supports focused testing activities when prompts become complex or require interaction with third-party systems.
- How deep it goes
- Defines four behavioral traits and demonstrates function-calling construction, while deferring advanced agent design.
- Examples
- Uses a test data assistant, multi-source analysis assistant, automation assistant, and exploratory testing assistant.
- What you can do after
- Build it from scratch as a focused LLM assistant that invokes tools, exchanges data, and supports a defined testing task.
Read how Software Testing with Generative AI explains AI agent →
Digital Twins in Action
An AI agent is a software system that uses an artificial-intelligence model to plan and execute sequences of actions toward a goal while interacting with tools and external state.
- How it is used
- Enables a digital twin to plan and execute multistep actions against changing physical-system state rather than only answer questions.
- How deep it goes
- Defines the agent loop and tools, illustrates a Python implementation, and discusses control boundaries and costs without detailing model internals.
- Examples
- Uses a solar-aware home climate agent, an inverter-query tool, and an HVAC setpoint adjustment.
- What you can do after
- Apply it via the Strands SDK to build a tool-using home agent while bounding its control authority.
Financial AI in Practice
A software system that perceives context, makes decisions, and takes actions to pursue specified goals.
- How it is used
- Deploys agents to move financial workflows from detecting risks to proposing proactive, context-sensitive interventions.
- How deep it goes
- Explains the agent loop, four architectural layers, cost filtering, and governance; treats MCP interaction as conceptual rather than runnable.
- Examples
- Uses a silent, at-risk investor, a fixed retention-email rule, and a posts-price-summary workflow to distinguish agent behaviors.
- What you can do after
- Build it from scratch as a tool-enabled retention workflow with ML filtering, LLM recommendations, human approval, and measured retention lift.
WordPress Power Toolkit: Harness AI to build next-level websites
An AI agent is an artificial system that perceives information, selects actions, and executes them to achieve programmed objectives.
- How it is used
- Helps evaluate and improve a website by visiting it, examining navigation and overall presentation, and suggesting improvements.
- How deep it goes
- Defines autonomy, goals, perception, decisions, and execution through a staged insurance-claim workflow without implementation details.
- Examples
- Uses automobile insurance claims and a site-review task that may compare competitors, assess visual appeal, and check plagiarism.
- What you can do after
- Evaluate/understand how to give a goal-oriented agent a short website-improvement request and anticipate its possible steps.
Read how WordPress Power Toolkit: Harness AI to build next-level websites explains AI agent →
Building Reliable AI Systems
A software system that interprets information, makes decisions, and takes actions to pursue goals.
- How it is used
- Turns language-model responses into tool-mediated actions and coordinated workflows, with reliability stakes beyond answer quality.
- How deep it goes
- Explains components, decision loops, memory, tool integration, and operational controls, then extends to harness isolation and multi-agent coordination.
- Examples
- Uses flight booking, calendar scheduling, and live weather queries to contrast generated answers, retrieval, and external action.
- What you can do after
- Apply it via named tools: use MCP for external tool access and LangGraph for shared-state, routed multi-agent workflows.
Build Python Web Apps with Streamlit
An AI agent is a software system that uses artificial intelligence to interpret context, choose actions, and carry out tasks toward a goal.
- How it is used
- Deploys AI in customer support to answer domain-specific questions and perform personalized actions through business systems.
- How deep it goes
- Explains reasoning and tool orchestration through ReAct and implements them with LangGraph nodes, conditional edges, and tool calls.
- Examples
- Demonstrates order tracking, order cancellation, tool calls, and a graph containing assistant and tools nodes.
- What you can do after
- Build it from scratch as a LangGraph support agent that calls tools and handles tracking or cancellation requests.
Read how Build Python Web Apps with Streamlit explains AI agent →
AI for Everyday IT
An AI agent is software that can pursue tasks toward objectives by interpreting information and taking actions with limited direct human intervention.
- How it is used
- Deploys AI agents for workplace administration, service-desk support, and increasingly complex specialized tasks.
- How deep it goes
- Explains initiative, system navigation, data analysis, and multi-step execution, but defers detailed implementation and agent internals.
- Examples
- Uses calendar management, email filtering, ticket triage, hardware troubleshooting, Salesforce Agentforce, Zendesk, Claude Desktop User, and Gemini 2 Flash.
- What you can do after
- Evaluate/understand where AI agents can automate support work while preserving verification, security, and human judgment.
Grokking AI Algorithms, Second Edition
An AI agent is a computational system that perceives context, makes decisions, and takes actions to pursue objectives.
- How it is used
- Uses agents to learn action sequences for goals in dynamic environments when the best plan is initially unknown.
- How deep it goes
- Defines agents, states, actions, environments, and rewards, then works through a simulator without developing learning algorithms in detail.
- Examples
- Uses a dog-learning analogy, a self-driving car in a parking lot, and reinforcement-learning game players.
- What you can do after
- Evaluate agent problems by identifying goals, states, actions, environments, rewards, and long-term consequences.
Read how Grokking AI Algorithms, Second Edition explains AI agent →
The Art of Code
An AI agent is a software system that performs tasks with some autonomy, typically by reasoning over inputs and invoking available tools or actions.
- How it is used
- Frames AI agents as developer tools for accelerating implementation while requiring direction and human supervision.
- How deep it goes
- Explains quality mirroring, correction loops, comprehension debt, and purity-oriented prompting without detailing agent internals.
- Examples
- Uses tangled or expressive codebases, repeated correction attempts, and a purity-focused code-review prompt as concrete examples.
- What you can do after
- Evaluate/understand generated code for design quality, hidden side effects, maintainability, and comprehension before accepting it.
Hugging Face in Action
An agent is a specialized AI system that autonomously performs tasks by understanding language, reasoning and planning, selecting tools, and delivering results.
- How it is used
- Deploys language models, planning, tools, and memory to handle ambiguous or multistep tasks through dynamic workflows.
- How deep it goes
- Explains agent roles, demonstrates a search agent, and extends implementation across smolagents, LangChain, and LangGraph.
- Examples
- Uses a weather agent, a DuckDuckGo search agent, hosted and local models, Python execution, and custom tools.
- What you can do after
- Build agents with smolagents, LangChain, or LangGraph and equip them with tools, execution logic, state, and memory.
First-Party Data Activation
A software system that uses artificial intelligence to perceive context, make decisions, and execute tasks toward goals with limited direct supervision.
- How it is used
- Deploys role-specific AI models to make decisions and execute coordinated marketing actions across integrated products.
- How deep it goes
- Defines roles, contexts, decisions, actions, integrations, and controls, then illustrates them through specialized marketing agents.
- Examples
- Uses Data, Audience, Journey, and Insights Agents to demonstrate data interpretation, segmentation, journey design, and performance recommendations.
- What you can do after
- Evaluate/understand where delegated AI actions can coordinate marketing workflows and where governance and human controls remain necessary.
Delivering Machine Learning Projects: From design to deployment
An AI system is a computational system that performs tasks associated with artificial intelligence, potentially incorporating machine learning models.
- How it is used
- Frames AI agents as deployed systems requiring user-facing controls, disclosure, validation, and failure governance.
- How deep it goes
- Defines design requirements and failure risks but stops short of model algorithms, training, or implementation internals.
- Examples
- Uses aircraft automation failure and smart-building signage as concrete design and governance examples.
- What you can do after
- Evaluate/understand an AI agent’s user-facing risks and translate them into disclosure, correction, and shutdown controls.
Read how Delivering Machine Learning Projects: From design to deployment explains AI agent →
Optimization Algorithms: AI techniques for design, planning, and control problems
An agent is an autonomous computational entity that senses information, interacts with other entities or its environment, and adapts its behavior.
- How it is used
- Frames autonomous interaction, adaptation, collective coordination, and goal-directed decision-making.
- How deep it goes
- Defines agent behavior through interaction and feedback, then distinguishes swarm, MDP, and bandit settings without detailing implementation internals.
- Examples
- Uses foraging, labor division, flocking, schooling, and reward-driven action selection as concrete agent examples.
- What you can do after
- Evaluate agent settings by comparing decentralized swarm behavior with reward-driven sequential or stateless decision-making.
The Complete Obsolete Guide to Generative AI
An AI agent is a software system that uses artificial-intelligence models to pursue goals by selecting and carrying out actions.
- How it is used
- Automates multilayer research by letting an LLM use search and other tools to answer a prompt.
- How deep it goes
- Demonstrates tool configuration and an execution trace, while defining reinforcement-learning agents separately through rewards and policies.
- Examples
- Uses a web-search question about an author’s titles and a two-model code generation and critique loop.
- What you can do after
- Apply it via LangChain by configuring search and math tools, initializing an agent, and inspecting its execution trace.
Read how The Complete Obsolete Guide to Generative AI explains AI agent →
Deep Learning with R, Third Edition
An intelligent agent is a system capable of adapting its behavior and selecting actions beyond merely executing hardcoded behavioral programs.
- How it is used
- Distinguishes adaptive intelligence from automation by focusing on behavior in novel situations.
- How deep it goes
- Explains adaptation, abstraction, reasoning, anticipation, and planning, but does not specify an agent implementation.
- Examples
- Uses thermostats, sorting programs, deep networks, organisms, and students facing new exam problems as comparisons.
- What you can do after
- Evaluate whether behavior reflects fixed automation or adaptation to unfamiliar situations.
Read how Deep Learning with R, Third Edition explains AI agent →
Automating API Delivery
A software agent is an automated software component that observes system state and performs actions to create, configure, or manage deployed applications.
- How it is used
- Continuously reconciles version-controlled API gateway configuration with deployed state to automate application management and recovery.
- How deep it goes
- Explains drift detection, redeployment, feedback loops, and rollback, but does not detail agent implementation.
- Examples
- Uses Kubernetes controllers, manual gateway changes, Git pull requests, automated redeployment, and git revert as concrete examples.
- What you can do after
- Operate/configure automated reconciliation, enforce repository-based deployment, and roll back API configuration with git revert.
Build Applications with Local AI Models on a Mac
An AI agent is a software system that perceives information, reasons about tasks, and takes actions to accomplish goals, including software-development tasks.
- How it is used
- Defines action-taking LLMs for tasks that may require choosing tools, reading results, and continuing through multiple steps.
- How deep it goes
- Explains tool choice and the evaluate-and-act loop, but stops short of connecting a model or implementing that loop.
- Examples
- Uses calculator and current-date tools in a Python listing, plus a calculator-choice scenario.
- What you can do after
- Evaluate/understand an agent’s tool choices and costs, then experiment with LangChain tools before building a full agent.
Read how Build Applications with Local AI Models on a Mac explains AI agent →
Building Distributed Applications that Work
An AI agent uses artificial intelligence to perceive context, make decisions, or perform tasks.
- How it is used
- Uses coding agents to inspect a .NET solution and incrementally adapt it for Aspire; also builds .NET agents with Semantic Kernel and large language models.
- How deep it goes
- Works through codebase analysis, AppHost proposals, implementation plans, and verification, but does not define agent internals.
- Examples
- Demonstrates Claude Code inspecting an online-shop solution, sketching its AppHost, and proposing a staged implementation plan.
- What you can do after
- Apply a coding agent to analyze a solution, propose Aspire changes, implement them incrementally, and verify the result before shipping.
Read how Building Distributed Applications that Work explains AI agent →
Related concepts
- language modeluses · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
- tooluses · A tool is an instrument or software capability that enables a person or system to perform a task, obtain information, or interact with resources.
- reinforcement learningused by · Reinforcement learning is a machine-learning paradigm in which an agent learns actions or a policy through interaction with an environment and evaluative feedback, such as rewards.
- conversational aiincludes · Conversational AI is artificial intelligence that enables software systems to interact with users through natural-language dialogue.
- agentic RAGused by · Agentic RAG is a retrieval-augmented generation approach in which an agent dynamically decides how and when to retrieve information, often iterating or using tools before generating a response.
- continuous improvementuses · Continuous improvement is an ongoing process of evaluating outcomes and making successive changes to improve a system, process, or result.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.