The Agentic AI Glossary

Claude Cowork is Anthropic's agentic desktop mode, available in the Claude desktop app, that lets Claude operate your computer — reading and writing files, running a sandboxed shell, driving apps, and connecting to your tools via MCP — instead of only chatting.

Agentic AI describes AI systems that are goal-directed rather than reactive — given an objective, they plan, take multiple steps, use tools, check their own work, and continue until the task is done, instead of just answering a single prompt.

The Model Context Protocol (MCP) is an open standard, introduced by Anthropic, that lets AI applications connect to external tools and data sources — like Google Drive, Slack, databases, and CRMs — through a consistent interface, so an AI agent can read from and act on the tools you already use.

Vibe coding is the practice of building software by describing what you want in plain language and letting an AI write and edit the code, so people can ship working apps and automations without traditional hands-on-keyboard programming.

Claude Code is Anthropic's agentic AI tool built specifically for software development. It works in the terminal and IDE, reading codebases, writing and refactoring code, running tests, and executing commands to help engineers ship software faster.

The OpenAI Codex app is a desktop application for macOS and Windows that acts as a command center for AI coding agents — running multiple agents in parallel, using background computer use to operate your apps, and automating routine engineering work with reusable skills.

ChatGPT Work is OpenAI's agentic desktop experience, built into a unified ChatGPT app for Mac and Windows, that can operate your computer, a built-in browser, and connected work apps to deliver finished work like spreadsheets, slide decks, and documents, rather than only chatting.

Computer use is an AI capability where the model operates a computer the way a person does — viewing the screen, moving the cursor, clicking, and typing — so it can use any application directly instead of relying on a special API for each one.

Human-in-the-loop is a design approach where an AI system pauses to get a person's approval before taking consequential actions, keeping a human in control of decisions that carry risk while still letting the AI do the heavy lifting.

A large language model (LLM) is an AI system trained on vast amounts of text to understand and generate human language, predicting likely continuations to answer questions, write, summarize, reason, and — when connected to tools — drive agentic workflows.

Claude is the family of large language models built by Anthropic, available through the Claude.ai web app, mobile apps, the Claude desktop app (home of Claude Cowork), Claude Code, and the Anthropic API.

Prompt engineering is the practice of writing clear, well-structured instructions that get an AI model to produce the result you want — specifying context, format, constraints, and examples so the output is accurate and useful.

AI workflow automation is using AI agents to carry out multi-step business processes end to end — gathering inputs, performing the work, and producing a finished output — turning recurring manual tasks like reporting, research, and intake into reviewable, on-demand results.

Answer Engine Optimization (AEO) is the practice of structuring content so search engines and AI assistants can lift a direct, self-contained answer out of it — for featured snippets, voice assistants, and AI Overviews — rather than just ranking a blue link.

Generative Engine Optimization (GEO) is the practice of optimizing content so it gets cited, summarized, or recommended by generative AI systems — ChatGPT, Claude, Perplexity, Google AI Overviews — instead of only ranking in a traditional list of blue links.