Vibe coding glossary

41 words you will run into while building with AI, explained without jargon.

A

Agent: An AI that can take several steps on its own — reading files, running commands, editing code — to reach a goal you describe, instead of answering one message at a time.
Agentic coding: Building software by giving an AI agent a task and letting it plan, write, test and fix the code itself while you review the results.
API: Short for application programming interface. A defined way for one piece of software to ask another for data or actions, like your app asking a weather service for today's forecast.
API key: A secret string that identifies your app to a service such as a payment or AI provider. Treat it like a password: never paste it into public code or chats.
Authentication: Checking who someone is, usually with a password, an email link or a Google sign-in. Often shortened to auth.
Authorization: Deciding what a signed-in person is allowed to do, such as whether they can edit a post or only read it.

Backend: The part of an app that runs on a server: the database, sign-in, and logic that should not run in a visitor's browser.
Bug: Anything in your app that does not behave the way you intended. Describing the bug clearly is the fastest way to get an AI to fix it.

Commit: A saved snapshot of your code at a moment in time, with a short message about what changed. Commits let you go back if something breaks.
Context window: How much text an AI model can consider at once, including your prompt, the files it reads and its own replies. Once it is full, older details get dropped or summarized.
Credits: The usage units many AI builders charge in. Each prompt or task spends some; plans differ in how many you get, so check each tool's own pricing page.

Database: Organized storage for your app's information, such as users, orders or posts, kept so it survives after a visitor closes the page.
Deploy: Putting your app on the internet so other people can use it. Many AI builders deploy with one button labelled Publish.
Domain: The human-friendly address of a website, like vibebuilder.space. You rent it from a registrar and point it at your deployed app.

Environment variable: A setting stored outside your code, such as an API key or database address, so secrets stay private and can differ between preview and production.

Frontend: Everything a visitor sees and clicks in their browser: pages, buttons, forms and layout.

Git: The most widely used tool for tracking changes to code over time. It records commits and lets several people work on the same project.
GitHub: A website that hosts Git repositories online, making it easy to back up code, share it and collaborate.

Hallucination: When an AI states something confidently that is not true, such as inventing a function, a fact or a link. Always check claims that matter.

IDE: Integrated development environment: a code editor with extra tools built in. Cursor and Windsurf are AI-first IDEs.
Iteration: Improving your app in small rounds: prompt, check the result, adjust, repeat. Small iterations are easier to review and undo.

LLM: Large language model. The kind of AI behind chat assistants and coding tools, trained on huge amounts of text to predict useful replies.

MCP: Model Context Protocol. An open standard that lets AI tools connect to outside data and services, like your files, a database or a project tracker, in a consistent way.
Migration: A scripted change to a database's structure, such as adding a table or a column, applied in order so every copy of the database stays in step.

No-code: Building apps by dragging and configuring blocks instead of writing code. Vibe coding differs because the AI writes real code you can keep.

Open source: Software whose code is published for anyone to read, use and change under a licence.

Preview vs production: Preview is a private test version of your app where you try changes. Production is the live version real visitors use. Check in preview, then publish.
Prompt: The instruction you give an AI. Clear prompts say what you want, who it is for and what done looks like.
Prompt engineering: The practice of writing and refining prompts so an AI produces reliable, useful results.

Rate limit: A cap on how often something can be done, like ten requests a minute. Services use it to prevent abuse; your app should use it on public forms.
Refactor: Reorganizing existing code so it is cleaner or easier to change, without changing what the app does.
Repo: Short for repository: the folder of code and its full change history, usually stored with Git.
RLS: Row Level Security. A database feature that decides, row by row, who may read or change data, so one user cannot see another user's private records.
Rollback: Returning your app or database to an earlier working version after a change goes wrong.
Rules file: A text file of standing instructions an AI tool reads on every task, such as coding style or things to avoid, so you do not repeat yourself.

SaaS: Software as a service: an app people use online and usually pay for by subscription instead of installing.
Server function: A piece of code that runs on the server when your page calls it, used for anything that needs secrets or should not be trusted to the browser.
Spec-driven development: Writing a clear description of what to build first — goals, screens, data, rules — and then having the AI build from that spec instead of improvising.

Token: A small chunk of text, often part of a word, that AI models read and write in. Usage and context limits are usually counted in tokens.

Vibe coding: Building software by describing what you want in plain language and letting AI write the code, then steering it by testing and giving feedback.

Webhook: An automatic message one service sends to your app when something happens, such as a payment succeeding.