 |
VOOZH | about |
|
VOOZH | about |
Claude Code has consolidated its position as the reference tool for agentic coding. The 2.1 line, launched in early 2026, is continuously updated: as of June 2026 it has reached release 2.1.160. Developed by Anthropic, this CLI (Command Line Interface) tool is transforming the work of developers worldwide, allowing them to automate complex tasks, navigate codebases of any size, and manage Git workflows in a completely natural way.
In this complete guide you will discover what Claude Code is, what is new in version 2.1, and why more and more developers consider it indispensable in their daily toolkit.
Here is a demo showing Claude Code working directly in the terminal:
As you can see, Claude Code receives instructions in natural language and translates them into concrete actions: modifies files, executes commands, navigates the codebase, and much more.
Claude Code is an agentic coding tool that lives directly in your terminal. Unlike classic AI assistants that require continuous copy-paste between interfaces, Claude Code integrates natively into your development environment.
Launched in February 2025 as an internal Anthropic research project, it has rapidly evolved into a very powerful tool. In 2025 alone, 176 updates were released, demonstrating an impressive development pace.
What Claude Code can do: Read entire codebases, plan complex modifications, write and debug code autonomously, execute shell commands, and work for hours on complex tasks - all through natural language commands.
The tech community has welcomed Claude Code with enthusiasm. Andrej Karpathy, AI legend, commented: "I have never felt so behind as a programmer... Clearly a very powerful alien tool has been delivered to us, except it comes without a manual."
Jaana Dogan, Senior Engineer at Google, stated: "I gave Claude Code a problem description, and it generated in an hour what we had built last year."
Version 2.1.0, released on January 7, 2026, brings numerous improvements that make Claude Code even more powerful and flexible. Let us see the main ones.
June 2026 update: since the January 2.1.0 release, the 2.1 line (now 2.1.160) has added dynamic workflows (orchestrating multiple sub-agents from a single script) and effort control, with Claude Opus 4.8 set to high effort by default for complex tasks.
Skills created or modified in the ~/.claude/skills or .claude/skills folders are now immediately available without having to restart the session. This greatly accelerates the development of custom workflows. For more, see the complete guide to Claude Code Skills.
You can now configure Claude response language with the language setting. For example, language: "italian" will make Claude respond in Italian. A feature highly requested by the international community.
For Vim lovers, version 2.1 introduces new movements:
; and , to repeat f/F/t/T movementsy operator for yank with yy/Yp/P for pasteiw, aw, i", a", i(, a( and others>> and << for indentationJ to join linesThe new /plan command allows you to enter plan mode directly from the prompt. Ideal for planning complex implementations before writing code.
This version includes numerous optimizations:
The name "Code" might be misleading. Claude Code is actually a general-purpose AI agent capable of performing almost any operation on your computer.
Combined with the Claude in Chrome extension (introduced in previous versions), it can control the browser, enabling automations that go well beyond writing code.
At Anthropic itself, Claude Code is used for:
Unlike traditional AI tools based on copy-paste, Claude Code introduces a completely different paradigm:
The AI environment integrates directly with your development environment. Claude Code executes npm commands, runs tests with Jest or pytest, uses Git for commits and branching, and starts development servers to verify changes.
A key element is the CLAUDE.md file: a special file that Claude automatically reads at the beginning of each conversation. Here you can document:
Before implementing, Claude Code can enter plan mode to explore the codebase, understand existing patterns, and propose an implementation plan. Only after your approval does it proceed with modifications.
Claude Code can launch specialized agents for parallel tasks. For example, an Explore agent (powered by Haiku) for quick codebase searches while the main agent works on something else.
The /review command provides subjective reviews that go beyond traditional linting:
Introduced in recent versions, the plugin system allows extending Claude Code with custom commands, agents, hooks, and MCP servers from the marketplace.
Version 2.1 introduces unified Ctrl+B to send both bash commands and agents to background. Background tasks now show clean completion messages instead of raw output.
Integration with Language Server Protocol has been improved with fixes for "no server available" errors during startup.
Hooks now support the once: true field and can be defined directly in the frontmatter of skills and slash commands.
To install Claude Code:
After installation, simply launch claude in the terminal to start an interactive session.
Requirement: A Claude account with API access or a Claude Pro/Team subscription is required.
No, it requires API credits or subscription. Claude Code uses Anthropic Claude models, so it needs an account with available API credits or a Claude Pro, Team, or Enterprise subscription. If you are looking for a free alternative, read how to use Claude models with OpenCode and Antigravity for free.
Yes, Claude Code is language-agnostic. It works with Python, JavaScript, TypeScript, Go, Rust, Java, PHP, Ruby, and practically any other language. It also understands configuration files, Docker, Kubernetes, and infrastructure.
Claude Code operates locally on your terminal. Code is sent to Anthropic servers for processing, so the same security considerations as using the Claude API apply. For sensitive projects, check company policies.
The agentic approach. While Cursor and Copilot are mainly autocomplete and inline assistants, Claude Code is an autonomous agent that can execute complex tasks end-to-end, navigate the filesystem, execute commands, and iterate on its own work.
Yes, with some limitations. It works natively on Windows, although some features are optimized for macOS and Linux. Version 2.1 includes specific Windows fixes, including improved WSL support for Claude in Chrome.
Absolutely yes. Many companies use Claude Code in production. Generated code belongs to you according to Anthropic terms of service.
Claude Code 2.1 represents a significant step forward in the world of AI-assisted coding. New features like skill hot-reload, performance improvements, and security fixes make it an even more reliable and powerful tool.
If you are a developer who wants to increase productivity without giving up control, Claude Code definitely deserves a try. The learning curve is minimal: you speak in natural language and Claude does the rest.
In summary: Claude Code is not just a coding assistant, it is a complete AI agent that is redefining how developers interact with code. Version 2.1 consolidates this vision with concrete improvements on all fronts.
Do you have questions about Claude Code or want to share your experience? Fill out the form at the bottom of the page!
Subscribe to the newsletter to receive new articles directly in your inbox.
Subscribe to the newsletter to receive new articles directly in your inbox.
3.4k readers worldwide, every Saturday