AI terms, in plain English.
Short, jargon-free definitions of the words you will meet when you start working with an AI agent.
The basics
The words you meet first: what models and agents are, and what you send them.
AI agent
An AI agent is software that works towards a goal on your behalf, using an AI model to decide the next step and tools to carry it out.
AI model
An AI model is a trained system that takes an input, such as text, an image or audio, and produces an output: an answer, a summary, a translation, a label.
Large language model
A large language model is an AI model trained on vast amounts of text to predict what comes next, which lets it write, summarise and answer questions.
Prompt
A prompt is the instruction or question you give an AI model, along with any material you include with it: an email to reply to, your notes, a document, an example of the result you want.
Token
A token is a small chunk of text, on average about three quarters of an English word, that AI models use to measure what they read and write.
Context window
A context window is the maximum amount of text a model can consider at once, including your instructions, the conversation, anything you paste in and its own reply.
How agents work
What turns a model into an agent that remembers, follows instructions and acts.
Agent engine
An agent engine, sometimes called a harness, is the software that turns a language model into an agent by adding memory, tools and the ability to act.
System prompt
A system prompt is a set of standing instructions given to an AI model before your conversation starts, shaping how it behaves in every reply.
AI memory
AI memory is information an agent stores between conversations, such as your preferences, the people you work with and the state of your projects, so it can carry context forward instead of starting from zero.
Agent skill
An agent skill is a saved, reusable procedure an AI agent applies when a task calls for it, bundling instructions, examples and sometimes scripts or access to a tool.
Scheduled task
A scheduled task is a job an AI agent runs automatically at a set time or on a repeating interval, without you asking for it each time.
Retrieval-augmented generation
Retrieval-augmented generation (RAG) is a technique in which an AI system first searches a set of documents for passages relevant to your question, then hands those passages to the model so it writes its answer from them.
Fine-tuning
Fine-tuning is extra training applied to an existing AI model using your own examples, so that it handles a specific task more consistently.
Safety & control
Keeping a person in charge, and what to watch for.
Human in the loop
Human in the loop is a design in which a person reviews, approves or can stop an AI system's action before that action is completed.
Guardrails
AI guardrails are the limits built into an AI system that restrict what it can say or do, reducing the chance of harmful or unwanted results.
Hallucination
An AI hallucination is false or invented information that an AI model produces and presents with the same confidence as correct information.
Connecting & hosting
How an agent reaches other software, and where it runs.
API
An API (application programming interface) is a defined set of requests one piece of software can make to another, and the responses it will get back.
API key
An API key is a secret string that identifies an application when it calls an API, working like a password for software.
Model Context Protocol
The Model Context Protocol (MCP) is an open standard that lets AI applications connect to external tools and data sources in a consistent way.
VPS
A virtual private server (VPS) is a rented, isolated slice of a larger physical machine that behaves like a computer of its own, with its own operating system, storage and network address.
Want a hand?
Talk to a real person about putting an agent to work in your business.