How to Learn Cursor
To learn Cursor, install the app, sign in, open a real project folder, and follow the quickstart: ask Agent to explain the codebase, make one small change, and review the diff. The docs describe Cursor as a coding agent for understanding a repo, planning features, fixing bugs, and reviewing changes.
Install and open a project
The install guide starts with an account: sign up free at cursor.com if you do not have one, download from cursor.com/download, and sign in when the app opens. The quickstart lists macOS 12 and later, Windows 10 and later, and on Linux an apt or dnf package. It prefers those packages over an AppImage because they add a desktop icon, automatic updates, and CLI tools.
Use File > Open Folder and choose a repository you can change. The first project guide starts there.
The first session
Open Agent with Cmd+I on Mac or Ctrl+I on Windows and Linux. The quickstart's first request is an explanation:
Explain this codebase. Point me to the main entry points, key modules, and anything I should read before making changes.
Agent searches the repo, reads relevant files, and summarizes how the project fits together. When that pass must stay read-only, switch to Ask mode by pressing Shift+Tab to cycle modes, or use the mode picker. Switch back to Agent mode when the next step should change code.
Then one small change. The quickstart's next prompt asks for three small, safe improvements, the tradeoffs, and a pause until you choose one. It names copywriting and small UI fixes as good first tasks. If you already know the edit, describe the result you want.
Edits apply as Agent works. Review them in the diff view and reject anything you do not want. Then ask Agent to run the checks the project already uses: tests, a type checker, lint, or a local build.
How Agent runs
The Agent overview splits an agent into instructions (the system prompt and rules), tools, and the model you pick. Tools search and read files, edit them, run shell commands, and more. There is no limit on how many tool calls one task can make.
Before significant changes, Agent saves a local checkpoint of the modified files, separate from Git. Restore it from the chat timeline to put the files back. The messages stay in the conversation. Use Git when you want the result kept.
Shell, MCP, and Fetch calls go through Run Modes in Settings > Agents > Approvals & Execution. Auto-review is what that page calls the safest useful setup for most people: allowlisted calls run, shell commands run in a sandbox when they can, and other calls go to a classifier that can ask you to approve.
Plan Mode for a larger change
Use Plan Mode when the task spans many files, has more than one reasonable approach, or should be approved before coding. Press Shift+Tab from the chat input to rotate to it. Cursor also suggests it when you type keywords that indicate a complex task.
Agent asks clarifying questions, researches the codebase, writes a plan you can edit, and waits until you click to build. Plans save in your home directory unless you click Save to workspace. If a build misses the goal, revert the changes, tighten the plan, and run it again.
Prompts and rules
Type @ to attach a file or folder (@auth.ts, @src/components/), a terminal, an earlier chat, a git diff, or the built-in browser. The prompting page says to use @ when you know the file, and to skip it when you do not. Agent searches on its own.
The model picker sits at the top of the input. On Mac, Cmd+/ cycles models, and the switch applies to the rest of that conversation. Faster models are for quick edits and routine tasks. More capable models are for complex reasoning and multi-file refactors. The current list is on Models & Pricing. When the context window gets close to full, Cursor compresses older turns into a summary.
Rules carry instructions into the next completion. The rules docs say models do not retain memory between completions, so an applied rule is included at the start of the context. Project rules are .mdc files in .cursor/rules. A plain .md file there is ignored. AGENTS.md, in the project root or a subdirectory, is the plain-markdown alternative. User rules under Customize → Rules apply to Agent across your projects. A rule can always apply, apply from its description or a glob, or apply only when you @-mention it. Keep it under 500 lines, and add one after Agent repeats a mistake. Rules do not change Tab. User rules do not apply to inline edit.
Tab and inline edit
Tab suggests code from recent edits, surrounding code, and linter errors, shown as gray text. Press Tab to accept, Escape to reject, or Cmd+Right Arrow (Ctrl+Right Arrow on Windows and Linux) to take the next word.
Inline edit is one selection. Select the code, press Cmd+K or Ctrl+K, type the change, and press Return. Opt+Return on Mac, or Alt+Return on Windows and Linux, asks a question about that selection instead. For anything wider, press Cmd+L or Ctrl+L. That opens Agent with the selection as context.
A course on that loop
The order is install, explain a real repo, one small edit, review the diff, then Plan Mode when the change is bigger than one file. Ailurn turns a prompt, a PDF, a GitHub repo, or a docs URL into a course you take in the same workspace. It does not watch YouTube videos. Pasting a link does not import a transcript. Free is planning chat, a full outline, and one lesson from a prompt. File attachments start on a paid plan.
Open the AI course builder and ask for a course that follows the Cursor quickstart: install, open a real project, make one small edit and review the diff, then practice the agent loop the docs describe (instructions, tools, and a model, with Plan Mode before a multi-file change).