Tool
A tool is an instrument or software capability that enables a person or system to perform a task, obtain information, or interact with resources.
Grokking AI Applications
An external callable capability that an AI application or agent can invoke to perform an operation or access information.
- How it is used
- Extends LLM applications with external data access and real-world actions.
- How deep it goes
- Defines tools and their descriptions, then traces agent calls through a calculator workflow without implementation internals.
- Examples
- Uses calculator calls and integrations with calendars, email, medical records, advertising dashboards, and CRM systems.
- What you can do after
- Apply it via Langflow by turning components or flows into tools for an agent.
Learn PowerShell Scripting in a Month of Lunches, Second Edition: Write and organize scripts and tools
A tool is a reusable, self-contained unit that performs a focused task and can be composed with other tools.
- How it is used
- Deploys focused PowerShell commands as reusable building blocks for automation workflows and business processes.
- How deep it goes
- Explains responsibility, parameters, pipeline composition, object output, naming, testing, modules, and separation from controllers.
- Examples
- Uses password-expiration pipelines, user provisioning, service changes, machine information, and help-desk interfaces as practical examples.
- What you can do after
- Build it from scratch as a scoped PowerShell function or module that accepts flexible input, returns objects, and composes with controllers.
AI Agents in Action, Second Edition
- How it is used
- Gives assistants and agents callable capabilities for retrieving context or performing tasks beyond text generation.
- How deep it goes
- Develops registration, tool selection, chaining, and failure guidance through examples, while deferring fuller production error handling.
- Examples
- Uses a research-source function, MCP search setup, and an order-lookup function with input validation and error handling.
- What you can do after
- Build focused function tools, connect MCP-hosted capabilities, and configure agents to use them with bounded failure handling.
Read how AI Agents in Action, Second Edition explains tool →
Data Analysis with LLMs
A tool is a callable software function whose interface is described in natural language so that a language model can invoke it.
- How it is used
- Lets agents choose and invoke functions to answer questions requiring database access, web search, calculation, or multiple processing steps.
- How deep it goes
- Defines tool invocation, traces database-tool calls, and shows how to expose a typed Python function with LangChain’s @tool decorator.
- Examples
- Uses a game-sales database and web search for a mixed-source question, then adds a currency-converter function.
- What you can do after
- Apply it via LangChain to load standard tools, create a SQL agent, or add a custom @tool function.
Spring AI in Action
A tool is a callable capability that allows an AI model or application to obtain data or perform an action through external functionality.
- How it is used
- Connects an LLM to real-time application data and actions when prompts require information unavailable from model training or RAG.
- How deep it goes
- Explains definitions, Java implementations, registration scope, provider interaction, message turns, and MCP exposure.
- Examples
- Demonstrates current time by city, board-game complexity from a database, and MCP methods for counting and filtering games.
- What you can do after
- Build it from scratch with Spring AI by defining, registering, testing, and tracing tools used during model interactions.
The Art of AI Product Development
A tool is a software capability that an AI system can invoke to retrieve information, communicate, execute operations, or affect an external system.
- How it is used
- Connects a language-model agent to information, operations, and external systems so it can perform tasks beyond generating responses.
- How deep it goes
- Defines tools as callable functions or APIs and discusses categories, selection, testing, and failure risks without providing implementation code.
- Examples
- Uses web research, email, roadmap updates, database queries, calculators, and a segmentation model to illustrate different tool roles.
- What you can do after
- Evaluate/understand which capabilities an agent needs, then assess tool reliability, selection errors, and the risks of write access.
Build AI-Enhanced Web Apps
A tool is an externally invocable software capability that an AI workflow can call to perform an operation or obtain information.
- How it is used
- Extends AI assistants with callable operations and information beyond text generation.
- How deep it goes
- Works through tool registration, client discovery, invocation, API-backed results, and failure handling in a local MCP integration.
- Examples
- Uses Wikipedia search in a LangChain.js agent and an API-backed joke tool in a Next.js chat application.
- What you can do after
- Build it from scratch as an MCP server tool, connect it to a Next.js assistant, and test it with MCP Inspector.
LLMs in Production
A tool is an external software capability or service that a language model can invoke to perform operations or access information.
- How it is used
- Extends LLM responses with current information, computation, retrieval, and actions beyond text generation.
- How deep it goes
- Explains tagged-call execution and prompt-based invocation, then demonstrates tool configuration; does not detail tool-training procedures.
- Examples
- Uses Guidance arithmetic functions and LangChain search, math, and Python execution tools.
- What you can do after
- Apply it via Guidance or LangChain to make callable capabilities available during model generation or agent execution.
AI Agents and Applications
- How it is used
- Provides agents with callable capabilities for retrieving information or taking actions during multi-step requests.
- How deep it goes
- Builds from a decorated retrieval function to combined local and remote tools, including inspection of tool calls.
- Examples
- Uses destination search over Wikivoyage content and a remote AccuWeather tool tested with a Penzance weather question.
- What you can do after
- Build it from scratch with a decorated function, then configure a LangGraph agent with local and MCP tools and verify its calls.
Build Python Web Apps with Streamlit
A tool is a callable software interface that an application or language model can invoke with structured parameters to perform an operation or retrieve information.
- How it is used
- Deploys callable functions or APIs so an agent can retrieve information and perform real-world actions during a conversation.
- How deep it goes
- Defines interfaces, examples, registration, and graph routing, while leaving function implementation details outside the focus.
- Examples
- Demonstrates user lookup, order lookup, status retrieval, and cancellation using static customer and order data.
- What you can do after
- Build it from scratch as documented functions, register it with an agent, and route model requests through tool execution.
Read how Build Python Web Apps with Streamlit explains tool →
AI Agents in Action
A tool is an external or internal capability that an agent invokes to perform an operation or obtain information.
- How it is used
- Extends agents with executable operations, API access, information retrieval, inspection, and coordination beyond ordinary conversation.
- How deep it goes
- Explains tool categories, attachment, documentation, and implementation examples, but defers broader reliability and maintenance discussion.
- Examples
- Demonstrates web retrieval, image generation, image inspection, assistant calls, and saving generated content to files.
- What you can do after
- Build custom actions, register them with an agent platform, attach them to workflows, and use them to complete or validate tasks.
Hugging Face in Action
A tool is an executable capability that an agent or language-model application can invoke to perform a task or access external functionality.
- How it is used
- Extends agents and language-model applications with executable access to search, APIs, files, computation, and other external functionality.
- How deep it goes
- Defines interfaces, wrapping, descriptions, execution flow, MCP registration, and testing, but stops short of implementation internals.
- Examples
- Demonstrates SerpAPI search, weather retrieval, temperature conversion, PDF and text access, and election lookup.
- What you can do after
- Build it from scratch by registering MCP functions, or apply it through LangChain and LangGraph agent APIs.
Digital Twins in Action
An agent tool is a callable function or external resource that an AI agent invokes to obtain information or affect an external system.
- How it is used
- Supplies live solar, forecast, and indoor data for an agent deciding when to run home cooling.
- How deep it goes
- Shows a documented Python tool and its API call, then places tool execution within an agent loop.
- Examples
- Uses an inverter query returning current power, daily energy, and a reading timestamp.
- What you can do after
- Build it from scratch with a Strands @tool-decorated function that documents its inputs and results.
Building LLM Applications with DSPy
An invocable software capability that performs an operation on behalf of an application or agent.
- How it is used
- Enables agents to carry out external operations that answer user requests, extending language-model reasoning into action.
- How deep it goes
- Defines tools by their role in agent execution and names common operations, but gives no interface or implementation details.
- Examples
- Names API calls, software execution, database updates, emails, purchases, and appointment scheduling as agent actions.
- What you can do after
- Understand how to pair language-model planning with executable actions in an agent.
Read how Building LLM Applications with DSPy explains tool →
Related concepts
- AI agentused by · An AI agent is a computational system that uses artificial intelligence to interpret information, select actions, and pursue goals through interaction with an environment.
- MCP resourcecontrasts with · An MCP resource is a URI-addressable source of data or content that a server exposes through the Model Context Protocol for clients to access.
- MCPconfigured by · The Model Context Protocol is an open standard protocol that defines how AI applications connect with external tools, data sources, and other contextual resources.
- functionkind of · A function is a callable program unit that performs an operation, may accept inputs, and can produce a result or side effect.
- ReActused by · ReAct is a DSPy module that implements an agent loop in which a language model selects tools and uses their results to guide subsequent steps.
- ReActused by · ReAct is an agent technique that iteratively interleaves reasoning, actions, and observations so results can guide subsequent actions.
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.