Describe what you want to be done in plain language, and Claude does it. Then, Claude plans the approach, uses the tools at its disposal, modifies files or performs the necessary actions, and, if applicable, verifies that it works.
For example:
Read my @shopping_list.xlsx and add a column to it with the prices from Amazon and Walmart.
Or,
I'm working on a machine learning model for pose estimation. I want you to check whether the preprocessing is being done correctly in my project.
Or, whatever comes to your mind ...
Yes, but Claude Code does things that those can't.
Regular AI chat (Claude, ChatGPT, Gemini) is like texting a very smart engineer. You describe a problem, they suggest a solution, you follow their instructions, report back what happened, and copy and paste again… It's a back-and-forth. Working with Claude Code is like giving the engineer your laptop and asking them to solve your problem. It can write code, run tests, read the results, debug errors, and adjust its approach, repeating the cycle on its own until something works, rather than giving you an answer (which might be totally off base) and waiting for you to test it.
Another major benefit of Claude Code is that it lets you create skills. Skills are saved instructions that you can name and trigger at any time. Instead of explaining a multi-step task from scratch each time, you define it once and reuse it forever. Skills let you capture that process once, so Claude executes it consistently without you having to remember or re-explain anything.
Another important benefit of Claude Code is the Model Context Protocol (MCP) tool creation (which is a big deal if you know some programming).
A paid Claude account (The free claude.ai plan currently does not include Claude Code access).
visit Official Documentation for more details or up to date instruction
Windows:
Open PowerShell and run the following command:
irm https://claude.ai/install.ps1 | iex
Or, use the following if you prefer to use Windows CMD:
curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd
Mac OS:
Open the Terminal application and run the following command:
curl -fsSL https://claude.ai/install.sh | bash
You need to create a project folder. Think of it as your home base! It's going to be where Claude Code will operate, and you'll keep your project's relevant data.
Pro Tip: Go to your project folder. If you have started writing a project description, prepared some code, or have other relevant files, you can move them here before initializing the project. Claude reads these files when initializing the project, making next steps easier.
To run Claude Code in our project, we need to go to that folder in the Terminal (on Mac) or PowerShell (on Windows). We have to do this first every time we want to work on the project in Claude Code:
cd "<the project folder address on your device>"
We can now initialize our project (we only need to do this once):
claude init
When initializing the project, Claude reads the files and creates a file called CLAUDE.md. This file defines the guiding context for a project, like a charter or manifesto, that Claude reads before getting involved. It typically includes things like coding conventions, big-picture notes, and any quirks or preferences specific to the project. You can, and probably should, modify this file to help Claude better understand the project and make it more closely align with yours.
/clear : Resets conversation
/compact : Free up context
/plan : Enter Plan Mode (no changes)
/model : Switch model (sonnet | opus | opusplan)
/voice : Toggle voice input (hold Space to speak, release to send)
/btw [question] : Ask side questions (no history pollution, no tools)
/rename [name] : Name or rename the current session
Shift+Tab Cycle permission modes
Esc × 2 Rewind (undo)
Ctrl+C Interrupt
Tab Autocomplete
Shift+Enter New line
Alt+T Toggle thinking
Space (hold) Voice input (requires /voice enabled)
Ctrl+D Exit
my-project/
├── .mcp.json # Project-scoped config
├── .gitignore # What Git ignores Claude ignores as well
├── mcp_server/
│ ├── pyproject.toml
│ ├── server.py # Your tools live here
│ └── .venv/ # API Keys and other secrets (should be added in .gitignore)
└── ... # Rest of your project