AI agent tools glossary
Plain-English definitions for the building blocks of AI coding agents - what each thing is, in one paragraph you can act on. Every term links to the catalog for it.
Skillselion's glossary defines the 21 building blocks of AI coding agents - MCP servers, agent skills, marketplaces, plugins and workflows - each in one plain-English paragraph, with a link to its live catalog of real, install-ranked tools.
MCP server
An MCP (Model Context Protocol) server is a small program that gives an AI coding agent like Claude Code, Codex, or Cursor a new capability - a tool, data source, or action it can call directly. Instead of copy-pasting context, the agent connects to the server and uses its tools live, over a standard protocol.
Model Context Protocol
The Model Context Protocol (MCP) is an open standard that lets AI coding agents like Claude Code, Cursor, and Codex connect to external tools, data, and actions through one uniform interface. Instead of bespoke integrations, an agent speaks a single protocol to any MCP server and calls its tools live - the plumbing behind every "MCP server."
Claude Code
Claude Code is Anthropic's agentic coding tool that runs in your terminal, IDE, and browser, letting Claude read, edit, and run code across a whole project. It extends through agent skills, MCP servers, plugins, and marketplaces - the ecosystem Skillselion catalogs - so you add capabilities without leaving your workflow.
Claude Code skill
A Claude Code skill is a reusable instruction set - a SKILL.md file plus optional scripts and references - that teaches Claude Code how to do one task well: a workflow, a convention, a domain. Skills install in a single command and load on demand, so the agent gains know-how without bloating its prompt.
agent skill
An agent skill is a reusable, installable instruction set - a SKILL.md file plus optional scripts and references - that teaches an AI coding agent how to perform a specific task well. Skills are loaded on demand, so an agent gains procedural know-how (a workflow, a convention, a domain) without bloating its base prompt.
marketplace
A marketplace is a registry an AI coding tool can add so it can install plugins and skills from it in one command. Adding a marketplace points the tool at a curated source; from there you install individual entries by name. It is the distribution layer for skills and plugins.
plugin
A plugin packages commands, sub-agents, or skills for an AI coding tool into a single installable unit, distributed through a marketplace. Where a skill teaches one task, a plugin can bundle several capabilities together and is added with one install command after its marketplace is registered.
workflow
A workflow is a multi-step, repeatable automation that chains an AI agent’s actions toward an outcome - for example scaffold, test, review, and open a PR. Workflows encode a sequence and its decision points, so a recurring engineering process runs consistently instead of being re-prompted each time.
Cursor
Cursor is an AI code editor, built on VS Code, that puts an AI agent inside your editor to write, edit, and refactor code across a project. It supports the Model Context Protocol, so you can extend it with MCP servers - the kind Skillselion catalogs - alongside agentic tools like Claude Code and Codex.
Codex
Codex is OpenAI's agentic coding tool that reads, edits, and runs code from your terminal, IDE, or the cloud, working through multi-step tasks on its own. Like Claude Code, it connects to MCP servers to gain new tools and data sources - part of the agentic-coding ecosystem Skillselion catalogs.
AI coding agent
An AI coding agent is a program that uses a large language model to read, write, and run code across a whole project with little step-by-step prompting - it plans and carries out multi-step tasks itself. Claude Code, Codex, and Cursor are AI coding agents; Skillselion catalogs the skills and MCP servers that extend them.
agentic coding
Agentic coding is building software by directing an AI agent that carries out multi-step engineering work - reading the codebase, editing files, running tests, and iterating toward a goal - rather than writing every line by hand. It is the workflow behind tools like Claude Code, Cursor, and Codex.
subagent
A subagent is a secondary AI agent that a coding agent spawns to handle a focused slice of a task - research, a parallel edit, a review - with its own context window, then reports back. Subagents let the main agent divide work and stay focused, and are often packaged inside skills or plugins.
vibe coding
Vibe coding is building software by describing what you want in plain language and letting an AI agent generate and refine the code, instead of writing it line by line. Coined by Andrej Karpathy in 2025, it leans on agentic tools like Claude Code and Cursor to turn intent into working code.
Idea phase
The Idea phase is where a developer turns a hunch into a validated opportunity using AI-coding tools - opportunity and competitor research, demand signals, and the skills, MCP servers and workflows that surface them. Skillselion ranks the strongest options for each job by real adoption, before you write a line of code.
Validate phase
The Validate phase is where a developer proves an idea is worth building before building it - scoping, prototyping and spikes, landing-page and waitlist tests, and pricing. Skillselion ranks the AI-coding skills, MCP servers and workflows for each de-risking step by real adoption.
Build phase
The Build phase is where a developer ships the actual product - UI and frontend, backend, data and payments, integrations, agent skills and templates, docs and project management. Skillselion ranks the AI-coding skills, MCP servers, marketplaces and workflows for each layer of the stack by real adoption.
Ship phase
The Ship phase is where a developer makes a release hold up under real users - testing and QA, code review, security, CI/CD and deploy, and performance. Skillselion ranks the AI-coding skills, MCP servers and workflows to catch what is easy to miss by hand, by real adoption.
Launch phase
The Launch phase is where a developer gets a shipped product found - distribution and launch channels, SEO for web search, app-store optimization, and AI-search / GEO visibility. Skillselion ranks the AI-coding skills, MCP servers and workflows to earn the first wave of attention, by real adoption.
Grow phase
The Grow phase is where a developer turns early users into retained, paying ones - analytics and insights, lifecycle and retention, content and marketing, and support and community. Skillselion ranks the AI-coding skills, MCP servers and workflows to compound traction, by real adoption.
Operate phase
The Operate phase is where a developer keeps a live product healthy and cheap to run - error tracking, monitoring and observability, infrastructure and cost, and iteration and experiments. Skillselion ranks the AI-coding skills, MCP servers and workflows to stay on top of production, by real adoption.