Products·Anthropic·Feb 2025

26. Claude Code Public Beta Launch

Agentic command-line coding tool that became Anthropic's fastest-growing product

Product Announcement

Summary

Anthropic launched Claude Code as a public research preview alongside Claude 3.7 Sonnet. Claude Code is an agentic command-line tool that enables developers to delegate entire coding tasks — debugging, refactoring, feature implementation, codebase navigation — directly from their terminal. Unlike chat-based coding assistants, Claude Code operates as an autonomous agent: it reads files, runs commands, writes code, and iterates on errors with minimal human intervention. Built on MCP for tool integration and CLAUDE.md for project context. Within months it became Anthropic's fastest-growing product, reaching $1B annualized run-rate revenue by early 2026.

Key Concepts

Terminal-native agentic coding: autonomous multi-step task execution

Claude Code runs in the developer's terminal with direct access to the file system and command line. Unlike chat interfaces where developers copy-paste code snippets, Claude Code reads project files, understands codebase structure, writes changes across multiple files, runs tests, and iterates on failures — all autonomously. The developer describes a goal; Claude Code executes the multi-step workflow to achieve it.

CLAUDE.md project context files

CLAUDE.md files provide persistent project context — coding conventions, architecture decisions, common patterns, and team-specific instructions. Unlike per-session context that vanishes, CLAUDE.md files travel with the repository, enabling Claude Code to understand project-specific requirements across sessions and team members. This became a widely-adopted pattern beyond Anthropic's ecosystem.

MCP integration for extensible tool access

Claude Code uses the Model Context Protocol to connect to external tools — databases, APIs, internal documentation, CI/CD systems. This means Claude Code isn't limited to file operations; it can query production databases, check deployment status, read Slack messages, or interact with any MCP-connected service. MCP turned Claude Code from a coding tool into a general-purpose developer agent.

Human-in-the-loop with permission boundaries

Claude Code asks for explicit permission before executing potentially destructive operations (git push, file deletion, running unknown scripts). Developers can configure trust levels from fully interactive to autonomous mode. This permission model balances agent autonomy with safety — critical for a tool operating directly on production codebases.