The complete AI glossary for 2026. Master key AI concepts from RAG and LLMs to Agentic Workflows.
AI vocabulary moves fast: agentic workflows, RAG, context engineering, and MCP went from research jargon to job-requirement language in under two years. This glossary defines the terms that actually appear in 2026 AI work: each entry written answer-first, with how-it-works mechanics, real-world examples, and the adjacent questions practitioners actually ask.
Every term links to the related concepts, tools, and industries where it applies, so you can move from a definition to the bigger picture in one click. Definitions are maintained against the current state of the field and dated for freshness.
For working fluency: LLM, RAG (retrieval-augmented generation), AI agents, context window, hallucination, fine-tuning, embeddings, and prompt injection. For the agentic era specifically: MCP, agentic workflows, tool use, and evaluation, they appear in most serious AI job descriptions and architecture discussions.
Generative AI is the broad category of models that create content; LLMs are the text-generating engines within it; agents wrap models in goals, tools, and action loops so they complete tasks rather than just answer. Each builds on the previous layer.
Entries are written against the 2026 state of the field and reviewed as terminology shifts: definitions include how concepts evolved (for example, prompt engineering's maturation into context engineering) rather than freezing 2023-era snapshots.