Claude Code

Claude Code is an agentic coding tool developed by the American artificial intelligence company Anthropic that performs software engineering work — such as reading and editing files, running shell commands, executing tests and investigating codebases — under the direction of a developer using natural-language instructions. It is one of the most widely used AI coding agents, alongside tools such as GitHub Copilot, Cursor and Devin, and is built on Anthropic's Claude family of large language models.^[1]

History

Anthropic released Claude Code as a research preview in February 2025, and made it generally available in May 2025. The tool was positioned as a shift from autocomplete-style code assistants toward agentic development, in which the model plans and executes multi-step engineering tasks with limited supervision. Through 2025 Anthropic expanded Claude Code beyond the original command-line interface: it added extensions for popular integrated development environments, integrations with source-control platforms, and in late 2025 a web-based version that lets users run coding agents in a browser without installing software. Enterprise versions added administrative controls, audit logging and compliance features aimed at regulated organisations.^[1][2]

Key Concepts and Technology

Claude Code operates as an agent that has access to a working directory and its tools. It can read and write files, run commands, search large codebases, and iterate on its own output — for example, implementing a feature, running the test suite and fixing the failures it introduced. A permission model lets developers review or automatically approve each class of action, and a plan mode makes the model propose a step-by-step approach before changing anything.

Two mechanisms shape how the tool behaves. Memory files such as CLAUDE.md store project conventions, architecture notes and preferred workflows that the model reads at the start of each session, giving teams a way to encode institutional knowledge. Subagents allow a main session to delegate bounded subtasks, such as documentation or test generation, to parallel model workers whose results are reviewed before being merged. Claude Code also supports plugins and hooks, and integrates with the Model Context Protocol (MCP), Anthropic's open standard for connecting models to external tools and data sources, which lets the agent call APIs, query databases and interact with services beyond the repository.^[1][3]

Applications and Impact

Claude Code is used for a broad range of engineering work: refactoring legacy code, writing and updating tests, reproducing and fixing bugs, generating documentation and exploring unfamiliar codebases. Development teams use it both interactively in the terminal and non-interactively in CI/CD pipelines, where agents can be dispatched on pull requests to propose changes. Anthropic reported rapid enterprise adoption through 2025, and the tool consistently appears near the top of public coding-agent benchmarks such as SWE-bench Verified, alongside competing products from Google, OpenAI and others.

The rise of Claude Code has also influenced the economics of software work. Access is included with Anthropic's Claude Pro and Max subscription plans, with heavier usage metered through the API, and the subscription model made agentic coding affordable for individual developers. Its adoption has contributed to debate about what programming skills matter: developers increasingly act as reviewers and directors of AI-generated code rather than authors of every line.^[2][3]

>See Also

• AI Agents
• Claude (Language Model)
• Model Context Protocol
• Vibe Coding
• Cursor (Code Editor)
• Devin