Claude Code review, pricing and verdict

Pick Claude Code when the job is a repo-wide change you want an agent to own, from the plan to the pull request. It is included in Claude Pro, Max, Team and Enterprise, never in Free, and its default model is Opus 5.5. For quick in-editor edits, Cursor or Copilot fit better.

Claude Code is Anthropic's coding agent for the terminal, desktop app, VS Code, JetBrains, the web and Slack. Its default model on paid plans is Claude Opus 5.5, with a native 1M-token window on Sonnet 5.5. It edits files, runs tests and opens pull requests, and since 7 August 2026 it runs in auto mode by default.

Maker: Anthropic · Autonomy: semi autonomous · Maturity: GA

Underlying models: Claude Opus 5.5, Claude Sonnet 5.5, Claude Fable 5.1

About Claude Code

Claude Code is Anthropic's coding agent, made generally available in May 2025 and now included in every paid Claude plan rather than sold as a separate product. It runs in a terminal, in the Claude desktop app, in VS Code and JetBrains IDEs, in a browser at claude.ai/code and from Slack, and it works on a real checkout: it reads files, runs shell commands and tests, edits across a repository, and opens pull requests through GitHub or GitLab. Anthropic's own Claude Code page names the main jobs as onboarding to an unfamiliar codebase, turning issues into pull requests, and running multi-hour refactors and migrations. HokAI lists the same product as a tool record, which carries the plan comparison; this agent record covers what the agent does and where it fits among other coding agents.

Which model it runs depends on the plan and on your settings. On Pro, Max, Team and Enterprise plans and on the Anthropic API, the default model is Claude Opus 5.5, and the sonnet alias resolves to Claude Sonnet 5.5, which has a native 1M-token context window. Claude Fable 5.1 is the most capable option in Claude Code, but it is never the default: you choose it with /model fable, it bills to usage credits on Pro, and Max plans draw on up to 50% of weekly limits for it. The opusplan setting uses Opus while planning and switches to Sonnet to write code, and a fallback chain can move an overloaded or safety-flagged request to another model for that turn. Opus 5.5 needs Claude Code v2.1.280 or later, so an old install has to run claude update first. The vendor is Anthropic.

Autonomy is semi-autonomous. Since 7 August 2026 Claude Code runs in auto mode by default on Pro, Max and Team plans, which lets it work for longer while still catching risky commands, and plan mode produces a plan for you to review before anything is edited. Projects, added to the desktop app on 17 September 2026, group related sessions so one person can supervise several agents at once, and self-hosted environments, in public beta since 6 August 2026, run sessions on your own infrastructure inside your network. It reaches outside systems through the Model Context Protocol, and agent teams let one session coordinate teammate sessions. Anthropic charges nothing separate for any of this, but there is no free tier, and every plan draws from the same usage pool as Claude chat, so a long agent session competes with your other work for the same allowance (see the Pricing tab for current rates).

Claude Code is the wrong pick in several cases. If most of your day is in an editor and you want inline completion and quick edits, Cursor or GitHub Copilot fit that rhythm better. With a ChatGPT subscription already on your card, OpenAI Codex comes bundled with it. If you want open-source code you can host and a choice of model provider, start with Hermes Agent, then Cline or OpenHands, which trade some polish for control, while Jules suits a single GitHub issue in a hosted sandbox. Some of these also use it: OpenClaw can run the Claude Code CLI as a native runtime. For a side-by-side with an editor-first rival, read Cursor Composer 2.5 vs. Claude Code, and to match an agent to your own stack, Smart Match asks about the work first.

Pricing

No free tier: the Free plan does not include Claude Code. Pro is $20/month ($17/month billed annually, $200 up front). Max 5x is $100/month and Max 20x is $200/month. Team Standard is $25/seat/month ($20 annual) and Team Premium is $125/seat/month ($100 annual); Team plans cover 2 to 150 people with no five-seat minimum. Enterprise adds usage at API rates. Fable models draw on usage credits on Pro and up to 50% of weekly limits on Max. Prices exclude tax and Anthropic can change them. Checked against Anthropic's pricing page on 2026-10-05. The [tool record](/hub/tools/claude-code) holds the full plan table.

Key Features

• Opus 5.5 by default: Paid plans and the Anthropic API start on Opus 5.5, and the sonnet alias points to Sonnet 5.5 with a native 1M-token window.
• Auto Mode: Auto mode is the starting setting for Pro, Max and Team users, so sessions can run longer with risky commands screened.
• Plan Mode and opusplan: Drafts a plan for review before editing, and the opusplan setting plans with Opus then writes code with Sonnet.
• Every surface: Runs in the terminal, desktop app, VS Code, JetBrains, claude.ai/code and Slack, all drawing on one shared usage pool.
• Projects and self-hosted runs: Projects group sessions for supervising several agents, and self-hosted environments (public beta since August 2026) keep sessions on your own compute.
• MCP and agent teams: Reaches databases, trackers and other outside tools over MCP, and lets a lead session hand work to teammate sessions.

Strengths

• Handles repo-wide work such as refactors, migrations and issue-to-pull-request jobs from one session, with plan mode before any edit.
• Ships on every surface from the terminal to Slack, so the same agent follows you between editor, browser and desktop.
• The default model is Anthropic’s top Opus tier and Sonnet has a native 1M-token window, so large codebases fit without manual file picking.
• Auto mode, projects and self-hosted environments give teams control over how long it runs unattended and where it runs.

Weaknesses

• No free tier: the Free plan excludes Claude Code, so trying it means paying for Pro or buying API credits.
• One usage pool is shared with Claude chat, with a rolling five-hour window and weekly caps, so long agent sessions use up your general allowance.
• Fable 5.1, the most capable model here, is never the default and is metered separately from your plan’s normal allowance.
• Newer models need a recent install, so an out-of-date version fails until you run claude update, and it is built around Claude models rather than a model-neutral setup.

Visit Claude Code Official Site