Claude Code

★ 135.4k · GitHub →

Anthropic's agentic coding tool that lives in your terminal.

Install

$ npm install -g @anthropic-ai/claude-code

What is Claude Code?

Claude Code is Anthropic's official coding agent. Instead of autocompleting lines inside an editor, it runs as a terminal session that can read your whole repository, plan multi-step changes, edit files, run tests and shell commands, and iterate until the task is done — with you approving each consequential action.

It is built around the Claude model family and speaks MCP (Model Context Protocol) natively, so it can be extended with external tools — databases, browsers, issue trackers — through the same protocol the rest of the agent ecosystem is converging on.

In practice it occupies the "senior pair programmer" slot: you describe intent at the level of a ticket or a code review comment, and it handles the mechanical spelunking across files.

Key capabilities

  • ▸ Full-repository awareness — reads and navigates the codebase rather than a single open file
  • ▸ Agentic loop: plan → edit → run tests → fix, with permission gates on shell commands
  • ▸ Native MCP client and server support for external tool integration
  • ▸ Git-native workflow: commits, branches, PR creation via the gh CLI
  • ▸ Hooks, slash commands, and skills for team-specific automation

When to use it

Pick Claude Code when the task spans many files and needs judgment — refactors, bug hunts, test-driven changes — and you want the strongest reasoning model in the loop. If you only need inline completion, an IDE plugin is cheaper; if you need a fully open-source stack, look at Aider or OpenHands.

Alternatives

Gemini CLI ★ 105.7k Aider ★ 46.9k OpenHands ★ 79.1k Cline ★ 64.2k

Frequently asked questions

Is Claude Code free to use?

No. Claude Code is included with Claude Pro/Max subscriptions and can also run pay-as-you-go against the Claude API. Costs scale with how much context the agent reads and how many iterations a task takes.

Does Claude Code work with MCP servers?

Yes — Claude Code is a first-class MCP client. You can attach MCP servers (databases, browsers, documentation sources like Context7) in its configuration, and the agent will discover and call their tools during a session.

How is Claude Code different from an IDE copilot?

A copilot suggests code where your cursor is; Claude Code executes tasks. It plans multi-step changes across a repository, runs commands and tests, and iterates on failures — closer to delegating to a colleague than to autocomplete.

Star count updated 2026-07-02 · refreshed weekly from the GitHub API · browse all tools →