GatwAI

Claude Code

Claude Code across terminal, IDE and cloud: repository work, plan costs, data policies and limitations with current sources and a dated hands-on report.

Provider sources checked

What does Claude Code do?

Claude Code is Anthropic's interface for development tasks with access to project files and tools. The agent can inspect code, prepare edits and run tests. Alongside the terminal, the provider documents desktop, IDE and web workflows. This differs from general Claude chat even though accounts and usage are connected. The practical distinction is working context: instead of copying code between chat and editor, you give the agent a scoped project task. Clearly described defects, migrations with acceptance criteria and documented maintenance are suitable candidates. Without meaningful tests and human review, extensive changes may merely look convincing.
Editorial research · 2026-11-14How we research →

Who is this tool for?

  • Developers working in terminals or IDEs.
  • Teams able to assess changes before merging.

Strengths

  • Project files, commands and tests can form one coherent task.
  • Terminal, IDE and hosted work support different working styles.
  • Commercial data policies are explicitly distinguished from consumer subscriptions.

Limitations and requirements

  • Long tasks consume varying allowances depending on context and model.
  • Passing tests establish only the behavior actually checked.
  • Tool access and external content require suitable permissions and review.

Who is behind the product?

Company
Anthropic
Headquarters
Not clearly dated by the primary sources reviewed.
Company founded
Not clearly dated by the primary sources reviewed.
Product launch
24 February 2025: limited research preview.

Anthropic's Claude 3.7 Sonnet announcement establishes Claude Code's introduction. That product date is not the company's founding date. Current functions and terms were checked separately against product, pricing and documentation pages.

Suggested ways to use it

Suggested workflows or verification steps. These are not hands-on tests; for unverified products they are not a recommendation to use the service.

Suggested: prepare a library migration

  1. Specify versions, affected modules and expected behavior.
  2. Review the migration plan before allowing scoped edits.
  3. Check tests and dependencies and document remaining work.

Goal: A reviewable migration state with explicit open issues.

Suggested: hosted maintenance task

  1. Deliberately select the repository and necessary network access.
  2. Delegate a bounded documentation or test problem.
  3. Inspect the branch and validation evidence locally before adoption.

Goal: Asynchronously prepared work with human approval.

Pricing, allowances and data

What should you check before choosing a plan?

The reviewed pricing page includes Claude Code from Pro: USD 20 monthly or USD 200 annually. Max starts at USD 100/month and Ultra at USD 200/month, with additional team and enterprise arrangements. Free Claude chat does not establish a free Claude Code tier. API authentication and token billing differ from subscription allowances.

What should you check about data and permissions?

For consumer plans, the model-improvement setting governs training use; documentation specifies five-year retention when enabled and 30 days when disabled. Commercial use is not used for training by default. Local session transcripts can be stored in plaintext. Enterprise retention arrangements require explicit qualification and configuration.

External reviews and available evidence

No GatwAI test. On 20 October 2025, Simon Willison described his own Claude Code web tasks, including a library benchmark.

Reported positives

  • He reports useful results and convenience from hosted execution.

Reported cautions

  • A single account using a supplied Max subscription, not an independent cost study or evaluation of today's version. He also discusses network boundaries.

These reports reflect the linked authors’ experiences. Test date, plan and task affect results; they do not establish a representative overall rating.

Product development and notable changes

  • Claude Code first announced

    Anthropic introduced the terminal agent as a limited research preview; today's plans and interfaces differ.

    Source for this change ↗

Which alternative fits which task?

Cursor →

For agent work in a dedicated development environment.

Frequently asked questions

Are Claude chat and Claude Code the same?

No. Claude Code works on files with development tools, although usage allowances may be shared.

Is a Pro subscription unlimited?

No. Tasks, context and model choice affect consumption; current usage rules apply.