Claude Code — A Coding Agent That Lives in the Terminal

Claude Code is an agentic coding tool from Anthropic. A developer asks for something in plain language and it reads the codebase, edits files across the project, runs tests and commands, and carries the work through to git commits and pull requests. It was first released on 24 February 2025 as a limited research preview, announced alongside Claude 3.7 Sonnet; the announcement said it let developers "delegate substantial engineering tasks to Claude directly from their terminal." It became generally available on 22 May 2025, the day Claude Opus 4 and Claude Sonnet 4 were announced.

If you want a one-line answer to "what is Claude Code", it is a tool for delegation rather than completion. Where GitHub Copilot-style completion suggests the next few lines inside an editor, Claude Code takes on planning, execution and checking as a single request. The official documentation defines it as an agentic coding tool that reads your codebase, edits files, runs commands and integrates with your development tools.

What it does

You start it by typing claude in a project directory and then hand it work in conversation. The main capabilities listed in the official documentation are these.

Where it runs has widened since the terminal-only start. As of October 2026 it runs in the terminal, in VS Code and JetBrains extensions, in a desktop app and in the browser at claude.ai/code, and the same CLAUDE.md and MCP settings apply on every surface.

From internal prototype to release

According to a September 2025 feature in The Pragmatic Engineer, the prototype was built by Boris Cherny, who joined Anthropic in September 2024. The first version could not read files or use bash and could do little more than control music through AppleScript; once he gave it filesystem and command tools, he said, it became really interesting. An internal version followed in November 2024, and about 20% of engineers used it on day one and about 50% by day five.

On the day of general availability Anthropic reported that Claude Opus 4 scored 72.5% on SWE-bench Verified, an evaluation built from real GitHub issues, and Sonnet 4 72.7%. Both are Anthropic's own measurements.

How far it spread

According to Anthropic, Claude Code reached $1 billion in annualised run-rate revenue in November 2025, about six months after public availability, and the company named Netflix, Spotify, KPMG, L'Oréal and Salesforce among its users. Its funding announcement of 12 February 2026 said the run-rate had passed $2.5 billion and more than doubled since the start of the year, that weekly active users had doubled since 1 January, and that enterprise use was more than half of Claude Code revenue. The same announcement cited an estimate by the research firm SemiAnalysis that 4% of all public GitHub commits were being authored by Claude Code. A run-rate is recent revenue annualised, not audited revenue.

Its makers changed how they worked too. In the same Pragmatic Engineer feature Cherny said that around 90% of Claude Code is written with Claude Code.

Availability in Japan and pricing

The general-availability announcement did not restrict the tool to particular countries. On 29 October 2025 Anthropic opened its first Asia-Pacific office, in Tokyo. The announcement said Rakuten was using Claude for autonomous coding projects and that the cloud integrator Classmethod reported Claude Code generating 99% of a recent project's codebase; that figure is Classmethod's own report.

Pricing below is from Anthropic's pricing page as of October 2026.

Pro and Max both have a five-hour session limit and a weekly limit. Team and Enterprise plans and pay-as-you-go API billing through the Anthropic Console also give access.

How the work changed

An analysis Anthropic published in April 2025, part of its Anthropic Economic Index, classed 79% of Claude Code conversations as automation, where the AI does the task with little human involvement, against 49% on the Claude.ai chat product. Startups were the main early adopters, and front-end work in JavaScript, TypeScript, HTML and CSS was prominent. Anthropic noted that the analysis did not measure code quality or actual productivity.

On productivity there is contrary evidence. In a randomised trial published by the research group METR in July 2025, 16 experienced open-source developers worked on 246 real issues, and tasks where AI tools were allowed (mainly Cursor with Claude 3.5/3.7 Sonnet) took 19% longer. The developers had expected a 24% speed-up and afterwards still believed they had been sped up by 20%. METR later described these results as outdated, and the sample covered only a narrow group of experienced developers. The distinction it drew between feeling faster and being faster has stayed in the debate.

Effects on junior jobs are covered on the page on AI and entry-level jobs.

The other side: misuse

In November 2025 Anthropic reported that GTG-1002, a group it assessed as Chinese state-sponsored, had used Claude Code as the core of an intrusion campaign. The account and the criticism of the report are on the GTG-1002 page. A tool that can run commands on a machine works for attackers as readily as for defenders.

The step before Claude Code, an AI that operates a screen, is on the page on Claude's computer use; the protocol it uses to reach outside tools is on the MCP page; and OpenAI's competing effort is on the OpenAI Codex page. The surrounding events are on the AI timeline and the AI and society timeline.