Glossary

Agentic search is a search mode where an AI agent breaks down a question into several searches and refines its answer, instead of returning a single list of results.

An AI agent is a software system that perceives its environment, makes decisions, and executes actions to reach a goal, with little or no human intervention.

An AI connector is the technical component that links a model or agent to a data source or external service, for example through MCP or an API.

The context window is the maximum amount of text, measured in tokens, a language model can take into account in a single request.

Fine-tuning means continuing the training of a foundation model on a specific dataset, to specialize it for a particular task or tone.

A foundation model is a large model trained on massive, general purpose data, designed to later be adapted to many specific tasks.

Generative AI refers to systems that create original text, images, code, or audio, rather than simply classifying or predicting existing data.

A hallucination is an AI generated answer that sounds plausible but is false, invented, or unverifiable.

A harness is the software layer wrapping an AI model. It manages context, tools, and steps, turning a raw model into an assistant that can act.

Inference is the moment a trained model generates an answer from new input, as opposed to the training phase.

An LLM is an AI model trained on massive amounts of text to predict and generate natural language. It powers most chatbots and assistants in use today.

Machine learning is a branch of AI where a system learns patterns from data instead of being explicitly programmed with each rule.

MCP is an open protocol that standardizes how a language model connects to external tools, files, and services.

An MCP client is the application or AI agent that connects to an MCP server to use the tools and data it exposes.

An MCP server exposes an application's data, files, or tools to a language model, following the MCP protocol.

NLG is the NLP subfield that produces natural language text from structured data or a given intent.

NLP covers the AI techniques that let a computer analyze, understand, or produce human language.

Prompt engineering is the practice of crafting precise instructions to steer a language model's output toward the intended result.

RAG pairs a language model with an external knowledge base queried in real time, so answers are grounded in specific sources instead of relying only on the model's memory.

A token is the basic unit a language model reads and generates, often a word, part of a word, or a punctuation mark.

A vector database stores vector embeddings and quickly retrieves the content closest in meaning to a given query.

A vector embedding is a numeric representation of a word, sentence, or image as a list of numbers, where mathematical closeness reflects closeness in meaning.