You do not need to learn everything. Start with this term.

An AI system that observes state, chooses next actions, and uses tools within defined permissions and limits to pursue a goal.

A reusable package of instructions and references that tells an agent when to activate, what to read, how to work, and what completion means.

Short for Model Context Protocol, an open protocol for connecting AI applications to external tools, data, and reusable prompts in a consistent way.

A product-specific installation and distribution unit that may bundle Skills, tool servers, connectors, or UI components.

Short for Large Language Model, a model trained on large amounts of text and other data to understand, generate, and transform language.

The amount of input and output a model can consider in one request, usually measured in tokens; excess content must be trimmed, summarized, or split.

A basic unit used to split text, which may be a character, part of a word, or punctuation; model limits and API costs are often measured in tokens.

Instructions and context given to a model, potentially including the task, data, examples, output format, and constraints—not merely one question.

The runtime stage where a trained model receives new input and produces a result, such as a chat response, image, or classification.

Short for Retrieval-Augmented Generation: retrieve relevant external information first, then give it to a model to generate an answer.

A numeric vector representation of text, images, or other data that lets systems compare semantic similarity by distance.

A data system optimized for storing and searching vectors, often used to retrieve document chunks for semantic search or RAG.

Short for Open Knowledge Format, an early open specification proposed by a Google Cloud team for portable agent knowledge made from Markdown, YAML frontmatter, directories, and links.

A browser API that uses the GPU to draw real-time 2D and 3D graphics in a canvas; Three.js provides a friendlier layer above it.

A JavaScript library built on browser graphics APIs such as WebGL, organizing real-time 3D around scenes, cameras, materials, and renderers.

A free, open-source 3D creation suite for modeling, materials, animation, simulation, and rendering, with Python automation for asset pipelines.

An open delivery format for real-time 3D that can describe meshes, materials, textures, rigs, and animation for the web or game engines.

A binary form of glTF that packages a scene and related binary data into one file for convenient deployment, downloading, and caching.

A visible 3D object made from geometry and material: geometry defines the shape, while material defines how the surface appears.

A small program executed on the GPU to determine vertex positions or pixel colors, forming the foundation of materials, lighting, and post-processing.

The component that sends a scene and camera through the graphics pipeline to draw the final image; Three.js commonly uses WebGLRenderer.

An approach where primary data and core functionality live on the user’s device and work offline; sync supports collaboration and backup instead of every action.

Running without the normal graphical interface so commands or code can automate work in batches, on servers, or in CI.

Short for continuous integration and continuous delivery or deployment: automated pipelines that repeatedly build, test, and release software.

An agreed interface that lets programs exchange data or call functionality, describing available operations, required inputs, and returned results.

A development kit for a platform or service, commonly including libraries, types, documentation, examples, and debugging tools.

A fast development style driven by describing intent in natural language and iterating on AI-generated code, while humans remain responsible for requirements, verification, and maintenance.

Metadata in the HTML head—such as title, description, and image—that tells social platforms how a shared link card should appear.

The preferred primary URL declared with a canonical link, used to consolidate duplicate or substantially similar page addresses for search engines.