Claude Code Review 2026: Pricing, Limits & Data Privacy

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 12, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Claude Code is Anthropic’s agentic coding tool. You describe a task in plain language, and it reads your repository, edits files across the codebase, runs commands and tests, and handles git, asking for permission according to the mode you pick. It started as a terminal program. It now also runs in VS Code (and Cursor), JetBrains IDEs, a desktop app, the browser at claude.ai/code, the Claude mobile apps and Slack.

It is built for developers ready to hand off multi-step work, such as fixing a bug end to end, writing tests or opening a pull request. It also suits teams that want the same agent in CI and in chat. It is tied to Anthropic’s Claude models: you cannot point it at OpenAI or Google models the way you can with model-agnostic agents.

This is an editorial research review based on Anthropic’s documentation, pricing pages and changelog, plus public reporting. We did not run our own benchmarks.

Quick verdict

Claude Code is a broad, mature coding agent. It covers the terminal, IDE, desktop, web and CI, with a deep customization system built on CLAUDE.md, skills, hooks, plugins and MCP. On paid Claude plans it defaults to Anthropic’s Opus 5.5 model with a 1M-token context window. The trade-offs are real. It only runs Claude models, and subscription usage limits are relative multipliers, not fixed numbers. Those limits have drawn repeated user complaints, and the product changes almost daily.

Pro: Runs in the terminal, VS Code and Cursor, JetBrains, desktop, browser, mobile, Slack and CI.
Pro: Deep customization shared across every surface: CLAUDE.md (or an existing AGENTS.md), skills, hooks, plugins and MCP servers.
Pro: Cloud sessions, Routines and pull request auto-fix keep working after you close your laptop, with no separate VM charge.
Pro: Mature admin and cost tooling for teams: spend limits, analytics, OpenTelemetry export and managed settings.
Con: Only runs Anthropic’s Claude models, served by Anthropic, Amazon Bedrock, Google Cloud or Microsoft Foundry.
Con: Subscription limits are relative (5x, 20x), not published quotas, are shared with Claude chat, and users report hitting them far sooner than expected.
Con: Not on the Free plan, and heavy API use gets expensive. Anthropic cites an average of $150–250 per developer per month in enterprise deployments.
Con: Closed source, with a release pace that changes behavior and defaults often.

Visit Claude Code

What is Claude Code?

Claude Code is a proprietary tool. Its license file says use is subject to Anthropic’s Commercial Terms of Service, so unlike the Codex CLI or Cline, it is not open source. Anthropic’s docs say every surface connects to the same engine, so a repository’s CLAUDE.md files, settings and MCP servers work in the terminal, IDE, desktop app and web alike.

You sign in with a Claude subscription (Pro, Max, Team or Enterprise) or a Claude Console API account. The CLI, VS Code and JetBrains integrations can also reach Claude models through Amazon Bedrock, Google Cloud’s Agent Platform or Microsoft Foundry. A one-line native installer covers macOS, Linux, WSL and Windows, and Homebrew, WinGet and Linux package managers work too. The JetBrains plugin needs the CLI installed separately, and the desktop app needs a paid subscription.

Key features

Agentic editing with permission modes

Claude Code can explore a codebase, plan, edit many files and run shell commands. Permission modes set how much it does without asking: Manual asks before most actions, acceptEdits lets edits through, and Plan mode proposes an approach without touching source files. In auto mode, a separate classifier model reviews actions instead of you. Since version 2.1.283, auto mode is the built-in starting mode for interactive terminal and VS Code sessions. /rewind restores both the conversation and the code to an earlier checkpoint.

Current Claude models and long context

On Pro, Max, Team, Enterprise and the Anthropic API, the default model is Opus 5.5. Aliases switch to Sonnet 5.5 or Haiku 5.5, and Haiku 5.5 became the default Haiku model on the Anthropic API on October 7, 2026. Anthropic’s Fable models must be selected explicitly and may bill to usage credits instead of plan limits. Newer models run with a 1M-token context window by default. Defaults differ on cloud providers.

Customization: CLAUDE.md, skills, hooks, plugins and MCP

A CLAUDE.md file (or an existing AGENTS.md) carries project rules into every session. Skills package repeatable workflows such as /review-pr. Hooks run shell commands before or after actions, such as formatting after each edit. Plugins install from marketplaces. MCP servers connect tools like Jira or Slack, and their tool definitions load on demand to save context.

Subagents, agent teams and background agents

Subagents work in their own context windows and report back summaries. Experimental agent teams, off by default, run several full instances under a lead agent; Anthropic warns they use roughly 7x more tokens in plan mode. A background agents view tracks parallel sessions, and the Agent SDK exposes the same tools for custom agents.

Cloud sessions, Routines and CI

Cloud sessions run in isolated, Anthropic-managed VMs and keep going while you are offline. Start them from the browser, mobile app, desktop app or claude --cloud, then pull them into your terminal with --teleport. Routines run on a schedule or trigger on API calls and GitHub events. Auto-fix watches a pull request and responds to CI failures and review comments. GitHub Actions, GitLab CI/CD and a GitHub Code Review integration cover automation. Cloud sessions are available on Pro, Max and Team, and for Enterprise users with premium or Chat + Claude Code seats.

Cost and admin controls

/usage breaks down plan usage by skills, subagents, plugins and MCP servers. Team and Enterprise admins get spend limits and an analytics dashboard, and Enterprise adds an Analytics API. OpenTelemetry export and managed settings work on every setup.

Who should use Claude Code

Claude Code fits developers who already pay for Claude Pro or Max and want one agent across terminal and IDE. It also fits teams that want a governed rollout with spend limits, analytics and managed settings, including organizations paying through AWS, Google Cloud or Azure. It is strongest on long, multi-file tasks.

Who should skip it

Skip Claude Code if you want to switch between model vendors or run local and open-weight models, if your policies require open-source tooling, or if you need a fixed, published quota for budgeting. It is also a weaker fit for teams on GitLab or Bitbucket who want cloud agents to open merge requests.

Strengths and limitations

Strengths

Breadth with one configuration. Every surface shares the same project rules, skills and MCP servers.
Customization depth. Hooks, skills and plugins let a team encode its conventions instead of re-prompting every session.
Long context by default. The 1M-token window on current defaults means less compaction on large repositories, though long sessions use more of your plan.
Local-to-cloud handoff. Teleport, Remote Control and the desktop handoff let one task move between laptop, browser and phone.
Transparent documentation. The changelog lists dated versions, and the cost docs publish typical spend figures.

Limitations

Single model vendor. Teams that want to route some work to OpenAI, Google or open-weight models need a different agent.
Opaque, shared usage limits. Pro and Max limits are multiples of each other, not published quotas, and are shared with Claude chat. DevOps.com reported in April 2026 that some Max subscribers had hit limits in as little as 19 minutes. Anthropic said it was investigating and that five-hour limits were being tightened during weekday peak hours.
Cost can climb. On API billing, Opus as the default model, long uncleared sessions, subagents and agent teams all increase spend. Anthropic’s docs name these as common causes of high bills.
Not open source. You cannot audit or fork the agent itself, unlike the Apache-2.0 Codex CLI or Cline.
Fast-moving defaults. The changelog shows nine versions between October 2 and October 9, 2026, including fixes for regressions introduced days earlier. With auto mode as the starting permission mode, teams that require human approval for every action must set that policy explicitly.
Cloud features favor GitHub. Cloning and pull requests in cloud sessions require GitHub. GitLab or Bitbucket repositories can be uploaded as a bundle but not pushed back. Zero Data Retention and IP allowlisting add further restrictions.

Pricing (as of October 2026)

Plan	Price	What you get
Free	$0	Claude chat only. Claude Code is not included.
Pro	$20/month billed monthly, or $17/month billed annually ($200 up front)	Claude Code on all surfaces. Usage shared with Claude chat; five-hour and weekly limits.
Max 5x	$100/month	5x Pro usage per five-hour session, higher output limits, priority access. Includes $100/month in API credits that cannot be spent in Claude Code.
Max 20x	$200/month	20x Pro usage per session, same extras, $200/month in API credits.
Team	Standard seat $20/month annually ($25 monthly). Premium seat $100/month annually ($125 monthly).	2 to 150 members. Premium seats get 5x standard-seat usage. Admin controls and spend limits.
Enterprise	$20/seat/month billed annually, plus usage at API rates	No fixed usage amount; admins set spend limits. Adds the Analytics API and an organization default model.
API or cloud provider	Pay per token	Claude Console, Amazon Bedrock, Google Cloud or Microsoft Foundry. No cloud sessions via third-party providers.

Subscribers can turn on usage credits to keep working past a limit at extra cost. For API billing, Anthropic’s docs put the enterprise average at about $13 per developer per active day. Prices exclude tax and can change, so confirm current terms on Anthropic’s pricing page before you buy.

See Claude Code plans

Prices verified on October 12, 2026 against the official Claude Code pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Claude Code is most useful when a task spans several files and steps. These scenarios follow Anthropic’s documentation as of October 2026 and assume you are comfortable working only with Claude models, since the tool does not run others.

Fixing a bug end to end from the terminal

A developer can describe a bug and have Claude Code explore the repository in Plan mode, which proposes an approach without touching source files, before letting it edit files, run the tests and commit. If a change goes wrong, /rewind restores both the conversation and the code to an earlier checkpoint. Auto mode, in which a classifier model reviews actions instead of you, has been the starting mode for interactive terminal and VS Code sessions since version 2.1.283, so choose a permission mode deliberately if you want to approve each action.

Encoding team conventions once

A team can keep project rules in CLAUDE.md (or an existing AGENTS.md), package repeatable workflows as skills, and add hooks such as running the formatter after every edit. Because every surface connects to the same engine, those rules apply in the terminal, IDE, desktop app and web alike. Behavior and defaults change often at Claude Code’s release pace, so use managed settings for the policies that matter and watch the changelog.

Keeping pull requests moving while you are offline

Teams on GitHub can run cloud sessions in Anthropic-managed VMs, use auto-fix to respond to CI failures and review comments on a pull request, and set Routines to run on a schedule or on GitHub events. GitHub Actions and GitLab CI/CD integrations cover pipeline automation. The caveat is platform support: cloud sessions clone repositories and open pull requests only on GitHub, while GitLab or Bitbucket repositories can be uploaded as a bundle but not pushed back.

A governed rollout through your cloud provider

Organizations that buy AI services through AWS, Google Cloud or Azure can run Claude Code against Claude models on Amazon Bedrock, Google Cloud or Microsoft Foundry and pay per token. OpenTelemetry export and managed settings work on every setup, and Anthropic’s Team and Enterprise plans add spend limits and an analytics dashboard. Cloud sessions are not available through third-party providers, and Anthropic’s cost documentation names long uncleared sessions, subagents and agent teams among the common causes of high API bills.

Investigating a large, unfamiliar codebase

Developers in large repositories can rely on the 1M-token context window that current default models use, which means less compaction on long tasks, and on subagents that research in their own context windows and report back summaries. This suits work such as tracing a bug across modules or mapping what a migration will touch before any code changes. Long sessions use more of a subscription or API budget, so clear context between unrelated tasks and check /usage to see what is consuming it.

Alternatives

OpenAI Codex: similar breadth across CLI, IDE and cloud, with an open-source CLI, and included in every ChatGPT plan, including Free and Go.
Cline: an open-source, model-agnostic agent for VS Code, JetBrains and the terminal that uses your own API keys.
Amp: a multi-model agent with cloud “orbs” that can also run on a Claude or ChatGPT subscription you already pay for.
Cursor: an AI-first editor, if you would rather work in a full IDE than a terminal agent.

Final verdict

Claude Code is a deep, well-documented agent with broad surface coverage and strong team controls, and it suits developers and organizations comfortable committing to Claude models. Before you standardize on it, weigh the relative, shared usage limits, the cost of heavy API use and the fast pace of change, and set your permission defaults on purpose. If model choice or open source matters more, look at the alternatives above.

Visit Claude Code

FAQ

Is Claude Code free?

No. Claude Code is not included in Anthropic’s Free plan; you need a paid Claude subscription (Pro, Max, Team or Enterprise) or pay-per-token billing through the Claude Console, Amazon Bedrock, Google Cloud or Microsoft Foundry. On subscriptions, Claude Code usage is shared with Claude chat, and you can turn on usage credits to keep working past a limit at extra cost.

Does Anthropic train on the code I send through Claude Code?

It depends on your account type. Anthropic’s data-usage documentation says that on consumer plans (Free, Pro and Max) it trains new models on your data, including Claude Code sessions, when the model-improvement setting is on, and you can change that setting at any time in Claude’s privacy controls. Under commercial terms (Team, Enterprise, the API and third-party cloud platforms), Anthropic does not train on code or prompts sent to Claude Code unless the customer opts in, for example through its Development Partner Program. Retention follows the same split: 30 days by default for commercial accounts, and either 30 days or five years on consumer plans depending on that setting.

Can Claude Code use GPT, Gemini or local models?

No. Claude Code runs only Anthropic’s Claude models, served by Anthropic, Amazon Bedrock, Google Cloud or Microsoft Foundry. The agent itself runs on your machine, but it is not designed to be pointed at OpenAI, Google or locally hosted open-weight models the way model-agnostic agents are. If model choice matters, Cline, which uses your own API keys, or the multi-model Amp is a closer fit.

Does Claude Code work in VS Code and JetBrains IDEs?

Yes. Besides the terminal, Claude Code runs as a VS Code extension (which also works in Cursor), a JetBrains plugin, a desktop app, the browser at claude.ai/code, the Claude mobile apps and Slack. The JetBrains plugin needs the CLI installed separately, and the desktop app requires a paid subscription. Project settings such as CLAUDE.md files and MCP servers carry across all of these surfaces.

How do Claude Code usage limits work?

On Pro and Max, usage is limited per five-hour session and per week, and Max plans are sold as multiples of Pro (5x or 20x per session) rather than as published quotas. The allowance is shared with Claude chat, and DevOps.com reported in April 2026 that some Max subscribers hit limits in as little as 19 minutes while Anthropic was tightening five-hour limits at weekday peak hours. On Team, Premium seats get 5x the usage of standard seats, Enterprise bills usage at API rates under admin spend limits, and /usage shows what is consuming your allowance.

Claude Code or OpenAI Codex?

Both cover terminal, IDE and cloud work, so the deciding factors are usually licensing, access and model preference. The Codex CLI is open source under Apache-2.0 and Codex is included in every ChatGPT plan, with narrower access on Free and Go, while Claude Code is proprietary, governed by Anthropic’s Commercial Terms, and needs a paid Claude plan or API account. In our assessment, teams should usually follow the model family they prefer and the subscription they already pay for; our OpenAI Codex review covers the other side.

Claude Code Review (2026): Anthropic’s Coding Agent, Limits and Pricing