Guide
What is Claude Code?
Harrowtech.ai · last checked 13 October 2026
Claude Code is a coding assistant from Anthropic that can read a project, change files and run commands for you. This guide explains what that means in practice, where you can use it and what you need before you start.
Harrowtech.ai is independent of Anthropic. Every outside fact below was read on Anthropic's own pages on 13 October 2026. Where a page said nothing, this guide says nothing.
The short answer
Anthropic's overview page calls Claude Code “an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools.” Another page says it is “an AI-powered coding assistant that helps you build features, fix bugs, and automate development tasks.”
“Agentic” is the key word. A chatbot answers a question and waits. Claude Code takes a task, then works through it in steps: it looks at your files, makes a change, runs something to check it, and carries on until the task is done or it needs you.
How it works, in three phases
Anthropic's How Claude Code works page says each task passes through three phases: gather context, take action, and verify results. They blend together. Claude may search files, edit one, run the tests, read the failure and edit again.
You are part of that loop. Anthropic says you can interrupt at any point to steer Claude, add context or ask for a different approach. Pressing Esc stops it straight away.
Two things do the work. The first is a Claude model, which reasons about the task. The second is a set of tools, which act. Anthropic groups the built-in tools like this:
| File operations | Read files, edit code, create new files, rename and reorganise. |
|---|---|
| Search | Find files by pattern, search content, explore a codebase. |
| Execution | Run shell commands, start servers, run tests, use git. |
| Web | Search the web, fetch documentation, look up error messages. |
| Code intelligence | See type errors after edits and jump to definitions. This needs a separate code intelligence plugin. |
What it can do
These are the uses Anthropic lists on its overview page.
- Everyday chores. Writing tests for untested code, fixing lint errors across a project, resolving merge conflicts, updating dependencies and writing release notes.
- Features and bugs. You describe what you want in plain language. For a bug, you paste the error or describe the symptom.
- Git work. It stages changes, writes commit messages, creates branches and opens pull requests.
- Connections to other tools. Through the Model Context Protocol (MCP), an open standard, it can reach things like Google Drive, Jira or Slack.
- Scripting. You can pipe text into it and run it without the interactive screen, for example
claude -p "your prompt"in a script. - Scheduled and parallel work. It can run on a schedule, and you can run several agents at once.
The “How it works” page adds that although it excels at coding, it can help with anything you can do from the command line, such as writing docs, running builds and researching topics.
What it can see
When you start Claude Code in a folder, Anthropic says it gains access to your project files, your terminal, your git state (branch, uncommitted changes, recent commits), your CLAUDE.md file, its own auto memory, and any extensions you have set up. Anthropic contrasts this with inline code assistants that only see the current file. For how CLAUDE.md and memory work, see Claude Code projects, CLAUDE.md and memory.
It can also edit and run things, so it asks first. Anthropic describes permission modes that you cycle with Shift+Tab: Auto, Manual, Accept edits and Plan. In Manual mode it asks before file edits and shell commands. Before it edits a file it snapshots the old contents, so you can press Esc twice to rewind. Anthropic is clear that this covers file edits only, not remote systems such as databases or deployments.
Where it runs
Anthropic's platforms page says Claude Code runs the same underlying engine everywhere, with each surface tuned to a different way of working.
| Terminal (CLI) | The full-featured version. Anthropic recommends it for terminal workflows, scripting and remote servers. |
|---|---|
| Desktop app | A standalone app with a visual diff viewer and parallel sessions. It includes Claude Code, so you do not install the CLI separately. A paid subscription is required. |
| VS Code and JetBrains | Extensions for working inside your editor. The JetBrains plugin needs the CLI installed separately. |
| Web | Runs in the cloud at claude.ai/code, with no local setup. Suited to long tasks that need little steering. |
| Mobile | The Claude app for iOS and Android starts and monitors cloud sessions, and can steer a session running on your computer through Remote Control. |
Anthropic says configuration, project memory and MCP servers are shared across the local surfaces. If you prefer to avoid the terminal, the desktop app vs terminal guide compares the two.
What you need to start
- A project folder to work in, and (for the terminal version) a terminal.
- An account. Anthropic's quickstart lists a Claude Pro, Max, Team or Enterprise subscription, a Claude Console account, or access through a supported cloud provider (Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry).
That is the whole list. The commands for each operating system are in How to install Claude Code, and what each plan costs is in Claude Code pricing. This guide does not repeat either.
Who it is for
Anthropic's own examples are aimed at people who write or maintain software: fixing bugs, writing tests, opening pull requests, setting up CI. The overview page also points to the Agent SDK for building your own agents with Claude Code's tools. Anthropic also publishes a terminal guide for people who have never used one, so a first-time terminal user is not shut out. Whether it suits your work is your call, and a small first task in a throwaway folder is a cheap way to find out.
Limits worth knowing
- Claude can be wrong. Anthropic's best-practices page says to give it a way to check its work, such as tests or a screenshot, and ask for evidence.
- Context is finite. Anthropic says performance drops as the context window fills, and that each new session starts with a fresh one.
- Instructions in
CLAUDE.mdare guidance, not enforcement. Anthropic points to hooks for things that must happen every time.
For habits that work around these limits, read Claude Code best practices.
What this guide does not cover
It does not cover installation, pricing, the full command list, or how Claude Code compares with other tools. Those have their own guides: commands and shortcuts and Claude Code alternatives. All guides are listed on the Guides page.
Sources
- Claude Code overview, read 13 October 2026
- How Claude Code works, read 13 October 2026
- Platforms and integrations, read 13 October 2026
- Quickstart, read 13 October 2026
- Best practices, read 13 October 2026
Harrowtech.ai's AI Setup service sets coding tools up with the rules and guardrails around how your developers work, from £720 (1 day), and coding tools training is a half-day from £360. See pricing, or get in touch.