What are Cursor, Copilot, and Claude Code?

What are Cursor, Copilot, and Claude Code?

Cursor, GitHub Copilot, and Claude Code are tools a developer uses while writing code. Each one puts a model to work inside the writing process: suggesting the next lines, answering questions about the codebase, or making whole changes on request. None of them ship inside the finished product. They sit on the same shelf as an editor, which is the shelf covered in the sorting lesson.

The three names also mark out the three shapes AI coding tools come in. The first shape is autocomplete: the tool watches what the developer types and suggests the rest of the line or the next block. The second is chat inside the editor: the developer types a question or a change request in plain language, with the code visible to both sides. The third is a coding agent: a tool that takes a task, plans the change, edits several files, runs the code, and reacts to what happens, while the developer reviews the result. Most of the products in this space now offer all three shapes, and each product started from a different one, which is why the names still carry different associations.

By 2026 these tools are near-universal among professional developers, and the last section comes back to what that fact does to reading them on a resume.

What is Cursor?

Cursor is a code editor with AI built into every part of it. It is a modified version of VS Code, the editor most developers already use, so it looks and behaves like a familiar workspace with a model attached. It autocompletes as the developer types, it has a chat panel that can see the open project, and it has an agent mode that makes multi-file changes on request.

Cursor is made by a company called Anysphere, and it grew fast for one plain reason: a developer could switch to it in an afternoon, because everything except the AI worked exactly like the editor they left. It became the shorthand name for AI-first editing, so when someone says a team "uses Cursor," they often mean the team writes code with heavy AI assistance inside the editor, whichever product is involved.

Cursor connects to models from Anthropic, OpenAI, and others, and the developer picks which one handles each request. The editor is the product; the intelligence is rented, the same arrangement the models lesson describes for AI products generally.

What is GitHub Copilot?

GitHub Copilot is an AI assistant that plugs into a developer's existing editor rather than replacing it. GitHub, the company that hosts most of the world's repositories, launched it in 2021, which makes it the oldest name on this page and the tool that introduced most developers to AI-assisted coding.

Copilot began as pure autocomplete. It read the code around the cursor and proposed the rest of the line, or the next several lines, in gray text the developer could accept with one key. That single interaction, suggestion and accept, was the entire product for its first years. Since then Copilot has added a chat mode and an agent mode, so the product now spans all three shapes.

The name is worth decoding. A copilot assists a pilot who stays in command, and the product was named to promise exactly that relationship: the developer flies, the tool suggests. On a resume, early Copilot adoption reads as being quick with new tooling, since almost everyone followed.

What is Claude Code, and how is it different from Claude?

Claude Code is a coding agent made by Anthropic that runs in the terminal, the typed command window from the first lesson. A developer describes a task in plain language, and the tool finds the relevant files, plans the change, edits the code, runs it, and reports back. The developer reviews the work the way they would review a colleague's.

The name causes a specific confusion worth settling. Anthropic is the company. Claude is the model the company trains. Claude Code is a tool built on top of that model. Three different things, three similar names, and resumes mention all three. "Experience with Claude" means building on the model through its API. "Uses Claude Code daily" means working with the tool. The same triple exists elsewhere: OpenAI makes the GPT models and a terminal tool called Codex, and Google pairs its Gemini models with a tool called Gemini CLI.

That abbreviation, CLI, appears constantly in this corner of the field. A CLI, or command-line interface, is a program operated by typing commands in the terminal instead of clicking in a window. Terminal-based coding agents took off because the terminal is where code gets run and tested anyway, so an agent living there can do the whole loop: change the code, run it, read the failure, fix it. The next lesson measures how far that loop goes with and without a person watching.

Is an AI coding tool the same kind of thing as React?

No, and the sorting machine from the shelves lesson shows why. React is a framework: it ships inside the product, its code runs every time a user opens the app, and choosing it shapes the codebase for years. Cursor and Claude Code are tools: they help while the work happens and leave nothing of themselves behind. A product built entirely with Claude Code contains no Claude Code, the same way a house built with a nail gun contains no nail gun.

The two shelves carry different weight on a resume. Framework experience is deep and slow to build, because the framework's habits and edge cases take years of real use to learn. Tool fluency is real and quick to acquire, closer to picking up a new editor than a new language. The last section puts a number on that.

There is one placement trap in this area. "Built with AI" can mean the candidate used these tools to write ordinary software, or it can mean the candidate built an AI product, software whose feature is a model, using techniques like RAG and agents. The first is a way of working. The second is a product domain. Job posts saying "AI-native" usually mean the second: the company's product depends on models, and the engineering work involves them daily.

What is vibe coding?

Vibe coding means building software by describing what you want and accepting what the AI produces, without reading the code closely. The phrase comes from a February 2025 post by the AI researcher Andrej Karpathy, who described giving in to the vibes: talking to the tool, accepting its output, and pasting error messages back in rather than studying the code. He was describing a real and useful mode for throwaway projects, and the phrase escaped into general use within weeks.

The mode works because current coding agents produce running software from plain-language description alone. For a prototype, a demo, or a one-off script, that is a genuine gift: the thing exists by the end of the afternoon, and nobody will maintain it. The trouble starts when unread code enters software a team keeps for years. Every line in a shared codebase eventually gets read, debugged, and changed by someone, and code nobody understood at writing time becomes technical debt with no author to consult. Teams that use AI heavily and well keep code review in the loop for exactly this reason, and a later lesson covers what an AI-heavy workflow leaves behind in the code.

How long does it take a developer to switch tools?

Days. The tools share the same working grammar: describe the task, read the suggestion, accept or redirect. The models underneath come from the same handful of labs, and a developer who has directed one AI tool has practiced the actual skill, which is specifying work clearly and judging the output. Moving from Copilot to Cursor, or from Cursor to Claude Code, resembles a salesperson moving between CRMs far more than it resembles learning a new language.

So the tool name on a resume describes a preference, roughly the way a preferred editor does. What separates developers in 2026 is how well they direct and review this work, at what scale, and with what judgment about when to trust it. Those are the questions the rest of this module takes up, starting from the module guide and the autonomy ladder one lesson over.