Claude Code

Delegate repository tasks from your terminal or IDE

by Anthropic

Overview

Claude Code delegates repository work to Claude: you describe the task and it plans, edits files, runs commands and tests across the project. It is terminal-first with official VS Code and JetBrains extensions, a desktop app and cloud sessions in the browser - the VS Code extension also installs into Cursor, and the JetBrains plugins need the CLI alongside them, so it does not require changing editors.

Which models you get varies with the plan, the provider and your organisation's policy rather than being one fixed list. On price, Pro is $20 a month or $200 a year, where the "about $17 a month" on the page is the annual equivalent rather than a monthly price; Max starts at $100 a month with 5x or 20x Pro usage; Team runs from 2 to 150 seats at $25 per seat a month, or $20 as the annual equivalent, for Standard, and $125 or $100 for Premium at 5x Standard usage; and Enterprise is $20 per seat a month annually plus usage metered at API rates, so the seat price is not unlimited model use.

Pro and Max share Claude Code usage with Claude chat across five-hour and weekly limits; verified Ultra has no five-hour limit. The API is separate pay-as-you-go, and reaching a limit means waiting for the reset, upgrading or switching on usage-based billing.

Local execution is still network inference: file edits and commands run on your machine while prompts and the relevant context go to Claude, and consumer and commercial accounts differ on training and retention - Pro and Max follow the model-improvement switch, with five-year retention when it is on and thirty days when it is off, while Team, Enterprise and the API are not trained on by default and retain for thirty days, and zero data retention is a separate arrangement for eligible Enterprise organisations. The self-hosted execution environment is a public beta for Team and Enterprise, enabled by an administrator, which puts the checkout, command execution and artefacts on your infrastructure while sessions still travel to Anthropic's control plane and model API; it cannot be run isolated, and those cloud sessions do not support switching to Bedrock, Google or Foundry or using zero data retention.

Editor's take

A strong candidate when you want repository-level delegation and you will review the diffs and run the tests yourself. What decides cost and fit: usage is shared with Claude chat rather than being its own coding allowance, the higher tiers buy more usage rather than unlimited use, and the execution model is not the data model - local execution still sends prompts and context to the model, consumer and commercial accounts have different training and retention rules, and the self-hosted runner beta moves execution onto your infrastructure without hosting the model.

Our reading of who this suits, based on the sources above. Opinion, offered to help you decide - not a lab result.

Key Features

Repository-level tasks: plan, edit, run commands and tests from a description
Terminal-first, with official VS Code and JetBrains extensions, a desktop app and browser cloud sessions
The VS Code extension installs into Cursor; the JetBrains plugins need the CLI as well
Usage shared with Claude chat under one subscription, with the API as separate pay-as-you-go
Pro $20/month or $200/year, Max from $100/month, Team seats from $25/month
Self-hosted execution environment in public beta for Team and Enterprise

Use Cases

Multi-file refactors with review
Fixing failing tests from a description
Delegating routine repository chores
Running agentic tasks in a cloud session

Pros

✓Runs in the terminal, an IDE or a browser session rather than requiring one environment
✓One subscription covers Claude chat and Claude Code, so usage pools in a single place
✓Published prices include an annual option for Pro and annual equivalents for team seats
✓Data handling is documented per account type, with a distinct zero-retention arrangement for eligible Enterprise organisations

Cons

✕Seat prices are not unlimited use: Max buys 5x or 20x Pro usage, Premium 5x Standard, and Enterprise adds usage metered at API rates on top of the seat
✕Pro and Max share five-hour and weekly limits with Claude chat; verified Ultra has no five-hour limit
✕Local execution is not offline: prompts and relevant context go to the model, and external MCP servers sit outside the zero-retention commitment
✕Training and retention differ by account: consumer plans follow the model-improvement switch, with five-year retention when on and thirty days when off, while commercial accounts are not trained on by default and retain for thirty days
✕The self-hosted execution environment is a Team/Enterprise beta an administrator has to enable, still connected to Anthropic's control plane and model API, and it cannot switch to Bedrock, Google or Foundry or use ZDR

Frequently Asked Questions

Where does Claude Code run?

It is terminal-first, with official VS Code and JetBrains extensions, a desktop app and cloud sessions in the browser. The VS Code extension can be installed into Cursor, and the JetBrains plugins need the CLI as well, so you do not have to change editors.

How much does it cost?

Pro is $20 a month or $200 a year - the "about $17 a month" shown on the page is the annual equivalent, not a monthly price. Max starts at $100 a month with 5x or 20x Pro usage, Team seats are $25 a month or $20 as the annual equivalent for Standard and $125 or $100 for Premium at 5x Standard usage, and Enterprise is $20 per seat a month annually plus usage metered at API rates. Prices exclude tax.

Do I get a separate coding allowance?

Pro and Max share Claude Code usage with Claude chat across five-hour and weekly limits. Verified Ultra has no five-hour limit. API access is separate pay-as-you-go. Hitting a limit means waiting for the reset, upgrading or enabling usage-based billing.

Does my code leave my machine?

Prompts and the relevant context go to the model, so local execution is not offline. Training and retention depend on the account: Pro and Max follow the model-improvement switch, with five-year retention when it is on and thirty days when it is off; Team, Enterprise and the API are not trained on by default and retain for thirty days; and zero data retention is a separate arrangement eligible Enterprise organisations can enable, with limits on the features it covers.

Can I self-host it?

There is a self-hosted execution environment in public beta for Team and Enterprise, enabled by an administrator, which puts the checkout, command execution and artefacts on your infrastructure. It is not a self-hosted Claude model: sessions still travel to Anthropic's control plane and model API, it cannot run isolated, and those cloud sessions do not support switching to Bedrock, Google or Foundry or using zero data retention.

Which regions may use Claude Code under its policy?

Anthropic's Supported Regions Policy explicitly includes Claude Code and permits its use only in the listed countries and regions. Restrictions consider physical location, entity headquarters or incorporation, and direct or indirect majority ownership or control by unsupported-region persons or entities. This records the official policy rather than an account or country reachability test; client entry points and account requirements are separately documented in the Claude Code overview. Claude Code is available in terminal, IDE, desktop app and browser surfaces. Most need a Claude subscription or Console account; Desktop Code requires a paid subscription, and native CLI installation is documented for macOS, Linux/WSL and Windows.

Source: Anthropic — supported regions policy · Claude Code - overview

Task guides for this product

T7 · Check whether code and prompts stay on your Mac

Key facts and where they come from

PriceNot established: The current pricing page gives Pro/Max/Team amounts and billing cadence, and explicitly labels the Enterprise annual seat amount US$. The consumer-plan currency-qualified purchase market and channel remain unestablished; annual equivalents are not no-commitment monthly prices.Source: Claude - pricing and plans
AvailabilityDocumented: Anthropic's Supported Regions Policy explicitly includes Claude Code and permits its use only in the listed countries and regions. Restrictions consider physical location, entity headquarters or incorporation, and direct or indirect majority ownership or control by unsupported-region persons or entities. This records the official policy rather than an account or country reachability test; client entry points and account requirements are separately documented in the Claude Code overview. Claude Code is available in terminal, IDE, desktop app and browser surfaces. Most need a Claude subscription or Console account; Desktop Code requires a paid subscription, and native CLI installation is documented for macOS, Linux/WSL and Windows.Source: Anthropic — supported regions policy · Claude Code - overview
Commercial useNot established: Claude Code applies Consumer Terms to Free/Pro/Max and Commercial Terms to Team/Enterprise/API, while third-party inference agreements can apply. Those terms assign output rights subject to conditions, and evaluations are personal non-commercial only; complete generated-code commercial conditions for every account, third-party provider and use case remain unestablished.Source: Claude Code - data usage
Training on your dataDocumented: Consumer Free, Pro and Max training follows the model-improvement setting. Commercial Team, Enterprise, API and other listed commercial accounts are not trained on under commercial terms unless they choose to supply data, such as through the partner program; retention varies by account. Historical scope read on 2026-09-23, not rechecked in full on 2026-10-15: Consumer Pro and Max plans follow the model-improvement switch, with five-year retention when it is on and thirty days when it is off; commercial Team, Enterprise and API accounts are not trained on by default and retain for thirty days, with third-party providers following their own policies.Source: Claude Code - data usage
API billingDocumented: Task length, model and features affect consumption, so there is no uniform number of coding tasks per month; at a limit the options are waiting for the reset, upgrading or enabling usage-based billing, and API access is separate from the subscription.Source: Claude Code - managing costs
Export formatsNot established: Claude Code CLI print mode explicitly supports text, JSON and stream-JSON responses. That bounded machine-response format does not establish a portable code/workspace export-package contract across terminal, IDE, desktop and web surfaces.Source: Claude Code - overview · Claude Code - self-hosted environments · Official claude-code reference
Usage limitsDocumented: Task length, model and features affect consumption, so there is no uniform number of coding tasks per month; at a limit the options are waiting for the reset, upgrading or enabling usage-based billing, and API access is separate from the subscription.Source: Claude Code - managing costs

Official sources

The vendor pages we read for the claims on this page, what each one states, and the day we read it.

Claude Code - overview ↗
Claude Code is available in terminal, IDE, desktop app and browser surfaces. Most need a Claude subscription or Console account; Desktop Code requires a paid subscription, and native CLI installation is documented for macOS, Linux/WSL and Windows. Historical registered scope read on 2026-10-15, not rechecked in full on 2026-10-15: Terminal-first, with an official VS Code extension, JetBrains plugins, a desktop app and cloud sessions in the browser; the VS Code extension also installs into Cursor and the JetBrains plugins require the CLI as well.
Read on October 15, 2026
Claude - pricing and plans ↗
Pro is listed at $20 billed monthly, with a separate annual payment. Claude Code is included from Pro and is not included on Free. Historical registered scope read on 2026-09-23, not rechecked in full on 2026-10-15: Pro $20 a month or $200 a year (the "about $17 a month" is the annual equivalent), Max from $100 a month with 5x or 20x Pro usage, Team 2-150 seats at $25 a month or $20 annual equivalent for Standard and $125 or $100 for Premium at 5x Standard usage, and Enterprise $20 per seat a month annually plus usage metered at API rates. Claude Code shares usage with Claude chat, and paid plans add weekly limits beyond the rolling five-hour window.
Read on October 15, 2026
Claude - using Claude Code with Pro, Max or Ultra ↗
Pro and Max share Claude Code and Claude chat usage across five-hour session and weekly limits. Verified Ultra has no five-hour session limit.
Read on November 2, 2026
Claude Code - managing costs ↗
Read on 2 November 2026: For Teams and Enterprise, Claude Code and chat share a per-seat allowance with five-hour and weekly windows. Historical scope read on 23 September 2026, not reread in full on 2 November: Task length, model and features affect consumption, so there is no uniform number of coding tasks per month; at a limit the options are waiting for the reset, upgrading or enabling usage-based billing, and API access is separate from the subscription.
Read on November 2, 2026
Claude Code - data usage ↗
Historical registered scope from 2026-09-23; the entire statement was not re-read on 2026-10-15: Consumer Pro and Max plans follow the model-improvement switch, with five-year retention when it is on and thirty days when it is off; commercial Team, Enterprise and API accounts are not trained on by default and retain for thirty days, with third-party providers following their own policies. Official body read on 2026-10-15, limited to this scope: Consumer Free/Pro/Max training follows the improvement setting. Commercial Team/Enterprise/API and other listed commercial accounts are not trained on under commercial terms unless they choose to provide data, such as through the partner program; retention varies by account. Consumer Free, Pro and Max training follows the model-improvement setting. Commercial Team, Enterprise, API and other listed commercial accounts are not trained on under commercial terms unless they choose to supply data, such as through the partner program; retention varies by account. Historical scope read on 2026-09-23, not rechecked in full on 2026-10-15: Consumer Pro and Max plans follow the model-improvement switch, with five-year retention when it is on and thirty days when it is off; commercial Team, Enterprise and API accounts are not trained on by default and retain for thirty days, with third-party providers following their own policies.
Read on October 15, 2026
Claude - zero data retention ↗
ZDR is enabled separately by eligible Enterprise organisations and is not part of a default plan; it covers requests made under the correct organisation credentials, with limits on the models and features it applies to, and cloud sessions, remote control, feedback and external MCP servers are among the cases it does not automatically cover.
Read on September 23, 2026
Claude Code - self-hosted environments ↗
Historical registered scope from 2026-09-23; the entire statement was not re-read on 2026-10-15: A public beta for Team and Enterprise that an administrator enables, placing the checkout, command execution and artefacts on enterprise infrastructure while session prompts, responses and tool results still travel to Anthropic services; it is not an isolated deployment, the Claude model is not self-hosted, and cloud sessions cannot switch to Bedrock, Google or Foundry or use ZDR. Official body read on 2026-10-15, limited to this scope: Team/Enterprise self-hosted environments are beta and off by default. They move session execution to customer compute; conversation/prompt/tool-result content still goes to Anthropic, and inference normally uses Anthropic or configured cloud providers. Self-hosted execution does not host the model on this Mac.
Read on October 15, 2026
Anthropic — supported regions policy ↗
Anthropic's Supported Regions Policy explicitly includes Claude Code and permits its use only in the listed countries and regions. Restrictions consider physical location, entity headquarters or incorporation, and direct or indirect majority ownership or control by unsupported-region persons or entities. This records the official policy rather than an account or country reachability test; client entry points and account requirements are separately documented in the Claude Code overview.
Read on October 15, 2026
Official claude-code reference ↗
The CLI reference documents --output-format text, json or stream-json for print mode. That response serialization format does not promise a complete downloadable code/workspace package across all Claude Code surfaces.
Read on October 15, 2026

Compare with other tools in Code Tools

Product	Maker	Pricing	Launched	Shared tags
Devin Desktop	Cognition	Freemium	2024-11-13	Coding, AI agent
Replit Agent	Replit	Freemium	2024-09	Coding, AI agent
Cursor	Cursor	Freemium	2023-03	Coding
GitHub Copilot	GitHub	Freemium	2021-10	Coding
Tabnine	Tabnine	Paid	2018-06	Coding
LM Studio	Element Labs, Inc.	Freemium	Launch date not verified	AI agent