Claude Code

Premium

By AI Bookmarks Editorial Team, Directory editors Last reviewed 2026-10-23

Visit Claude Code Official site · opens in a new tab

Anthropic terminal coding agent that reads a repository, edits files and runs commands, asking before risky steps.

Quick facts

Pricing	Premium
Free tier	No — requires a paid Anthropic plan or API billing
Platforms	Terminal, VS Code, JetBrains
Developer	Anthropic
Website	https://claude.ai

What Claude Code is

Claude Code is a command-line coding agent from Anthropic. You run it inside a project directory, describe a task, and it reads the relevant files, makes edits, runs tests and shows the diff, asking for approval before actions that change the repository depending on how it is configured. Because it works from the terminal, it fits existing scripts and CI rather than requiring a new editor, and it can also be invoked from an IDE extension. It handles multi-file changes and can be pointed at a specific directory or the whole repository. The limits are oversight and cost: long sessions consume a substantial amount of the model allowance, the agent follows instructions literally in ways that can surprise, and changes still need review before merge. A memory file at the project root records conventions, so the agent does not need the same instructions every session. Hooks allow scripts to run before or after tool calls, which is how teams enforce formatting.

Best for

Refactors that touch several files at once
Scripting a coding task into an existing terminal workflow
Reviewing a diff rather than typing every line

Not ideal for

Anyone unwilling to review generated changes
Work that must stay entirely offline

Pros

Works from the terminal, so it fits existing tooling
Runs tests and reports results before you review
Approval prompts keep destructive steps visible

Cons

Consumes a large share of the model allowance
Literal instruction-following can surprise
No free tier

Pricing in detail

Claude Code has no free tier: it is used through a paid Anthropic subscription with included usage, or billed per token through the API. Because each task can involve many model calls, a long refactoring session costs far more than a chat conversation. Teams should watch usage during the first weeks rather than assume the included allowance covers a full working day.