The AI design vocabulary,
defined plainly.

An MCP (Model Context Protocol) server exposes tools, data, and prompts that an AI agent can discover and call over a standard protocol. Any MCP-compatible client — Claude, Cursor, Claude Code — connects to the server and gains its capabilities as native tools, with no custom integration code.

An AI agent is a system that uses a language model to pursue a goal across multiple steps — planning, calling tools, observing results, and adjusting — rather than producing a single response. Agents can read data, run code, generate assets, and act in software on a user’s behalf.

Context engineering is the practice of deciding what information a model sees at inference time — which documents, memories, tools, and instructions to place in the context window — so it produces accurate, grounded output. It has largely superseded "prompt engineering" as models and context windows have grown.

A company brain is a queryable knowledge layer that holds an organization’s context — brand, documents, policies, product data — and feeds it to AI agents so their output is grounded in how the company actually works. It turns scattered institutional knowledge into agent-usable skills.

Agent memory is how an AI agent retains information across turns or sessions — facts, past actions, user preferences — instead of starting blank each time. It ranges from short-term (the current context window) to long-term (a persistent store the agent recalls from when relevant).

Generative UI is interface that an AI generates or assembles on the fly in response to a task, rather than being hand-built in advance. Instead of navigating fixed screens, a user states a goal and the system produces the controls, layout, or visualization that fits the moment.

An editable SVG is vector artwork stored as real, structured path code — shapes, groups, gradients — that can be re-colored, re-shaped, and scaled to any size with no quality loss. It contrasts with a flattened or rasterized export, which is just pixels of a picture and can’t be meaningfully edited.

Vector graphics describe an image as math (paths, curves, fills) so they scale infinitely with no pixelation; raster graphics store a fixed grid of pixels that blurs when enlarged. Vectors suit logos, icons, and anything printed at multiple sizes; raster suits photographs.

An MCP client is the AI application that connects to MCP servers and calls their tools — Claude Desktop, Claude Code, Cursor, or any app that speaks the Model Context Protocol. The client is where the model runs; it discovers a server’s tools and lets the model invoke them mid-task.

Agent orchestration is coordinating multiple AI agents — or multiple steps of one agent — toward a goal: routing each task to the right agent, running them in parallel or sequence, and combining their results. It is how a system does work too big or varied for a single model call.

Headless generation is producing an asset — an image, a vector, a video — from an API or agent with no user interface or browser involved: prompt in, file out. It lets automated systems and agents create media server-side, at scale, as part of a pipeline.

Vectorization (image tracing) converts a raster image — pixels — into vector paths: math-described shapes that scale to any size without blur. It is how a PNG or photo becomes an editable, resolution-independent SVG suitable for logos, cutting, and print.

Brand DNA is the codified essence of a brand — its palette, typography, tone, and visual rules — stored so that people and AI agents produce on-brand work automatically. It turns a static brand guideline into an active constraint every generation obeys.

Lottie is a JSON-based format for vector animation — small, scalable, and rendered natively on web and mobile. Exported from design tools, a Lottie plays crisp motion (an animated logo, icon, or loader) at any size without the file weight of a video.

Prompt engineering is crafting the input to a model to get a better output — choosing wording, structure, examples, and constraints. For image and vector generation it is how you steer subject, style, composition, and format; as models and context windows grow, it shades into context engineering.

A design system is the shared library of components, styles, and rules a team designs and builds with — colors, type, spacing, icons, and reusable parts — so work stays consistent and fast. In the AI era it becomes the constraint that keeps human- and agent-generated output coherent.