What is Claude Code?
Claude Code is Anthropic's agentic coding tool: it reads your codebase, edits files across it, runs commands, and works with git — driven by plain-language instructions rather than autocomplete. It started as a terminal program in February 2025 and now runs in the terminal, VS Code and JetBrains, a desktop app and the browser, all on one engine that shares your project instructions and settings between them.
- Made by
- Anthropic makers of Claude
- Introduced
- Feb 2025 as a research preview
- Category
- Coding agent not an IDE, not autocomplete
- Runs in
- 4+ surfaces terminal · IDE · desktop · web
- Costs from
- $20 / mo · no free tier
- Source
- Proprietary public repo ≠ source code
- Extends via
- MCP + skills · hooks · subagents
- Best for
- Multi-file work in a codebase you already have
Anthropic's own one-line definition, and the date it first appeared.
The docs open with the definition we quote above: an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools, available in the terminal, IDE, desktop app and browser. That phrasing is the vendor's, not ours.
Its first public appearance is dated. Anthropic introduced it on 24 February 2025, alongside Claude 3.7 Sonnet, describing it then as "a limited research preview" that "enables developers to delegate substantial engineering tasks to Claude directly from their terminal". Everything since — the IDE extensions, the desktop app, the browser version — has been built around that same terminal core.
One correction worth making, because search results get it wrong: Claude Code is not open source. The public anthropics/claude-code repository holds issues, a changelog, plugins and examples — the application's source code is not published there.
What it actually does.
The useful distinction is between a completion tool and an agent. Autocomplete proposes the next few lines and waits. Claude Code takes an instruction — "write tests for the auth module, run them, and fix any failures" — then finds the files itself, makes edits across several of them, runs the tests, reads what broke, and iterates.
The same loop covers the chores around the code: staging changes, writing commit messages, opening branches and pull requests, resolving merge conflicts, updating dependencies. In CI it can review pull requests and triage issues through GitHub Actions or GitLab.
It began as a terminal program and still is one at heart, but the same engine now drives a VS Code extension, a JetBrains plugin, a desktop app, a browser version at claude.ai/code, and sessions you can reach from a phone. Anthropic's framing is that these are surfaces over one product: your CLAUDE.md project instructions, settings and MCP servers apply to all of them.
That matters when comparing it to editor-bound tools. Cursor is an editor you work inside; Claude Code is an agent you can attach to whatever you already use — including Cursor, via the extension.
Three extension points do most of the customising work. MCP (Model Context Protocol) connects it to outside systems — a ticket tracker, a design doc, an internal API. Skills package a repeatable workflow your team can invoke by name. Hooks run your own shell commands before or after its actions, so formatting or linting happens automatically.
Above that sit subagents and agent teams, where several sessions split a task and a lead session merges the results, and the Agent SDK for building your own tools on the same machinery.
Four things it gets mistaken for, each answered on its own page:
- Not an IDE. It has no editor of its own — it works on the files you already have, wherever you keep them.
- Not free. A paid Claude plan or pre-paid API credits are required; there is no free tier.
- Not open source, despite what several search results claim.
- Not usable offline — every turn is a call to a model over the network, as we measured.
Tools people weigh it against.
All three are agentic rather than autocomplete, and each makes a different bet about where the agent lives — in your editor, in your terminal, or in your CI. We've written each pair up separately with pricing and benchmarks.
Frequently asked.
Quick follow-ups people search after this question.
Q · 01 What does Claude Code actually do? +
Q · 02 Is Claude Code the same thing as Claude? +
Q · 03 Does it work inside VS Code? +
CLAUDE.md, settings and MCP servers carry over from the terminal.Q · 04 What does it cost? +
Q · 05 Is Claude Code open source? +
anthropics/claude-code repository on GitHub is a public home for issues, the changelog, plugins and examples — the application source is not published. Several search results and AI summaries state otherwise; they are wrong.