Tool profile
Claude Code: what it is, what it costs, where it breaks
Claude Code is an agentic coding tool made by Anthropic that reads a codebase, edits files, runs shell commands and works with git from a terminal, IDE extension, desktop app or browser. Developers hand it a task in plain language and review the changes it makes, rather than typing code line by line.
What to know first
Claude Code is not sold as a separate product with its own seat. Access comes with a paid Claude subscription, a Team or Enterprise seat, or Anthropic Console credits, and the usage you get depends on that account. The same engine and the same project configuration files run on every surface, so a CLAUDE.md written for the terminal also applies in VS Code or the web. Training on your code depends on account type.
Claude Code at a glance
| Maker | Anthropic, United States |
|---|---|
| Free plan | No |
| Entry price | $20 per month (Claude Pro, or $17 billed annually) |
| API access | Yes |
| Category | Best AI Coding Assistants in 2026, rank 3 of 9 |
| Website | claude.com/product/claude-code |
Who Claude Code is best for
Developers who delegate whole tasks
Writing tests for a module, fixing lint errors across a project or resolving merge conflicts suits a tool that works through a task and returns a result, with you reviewing the diff afterward.
Teams that automate with scripts and CI
The command line accepts piped input and a non-interactive print mode, and GitHub Actions and GitLab CI/CD are documented, so reviews and issue triage can run without anyone at a keyboard.
Teams with a standing cloud contract
The terminal, VS Code and JetBrains versions can run against third-party providers, so an organization that already buys model access elsewhere can keep that arrangement and billing.
People who move between devices
Sessions can start on the web or the mobile app and be pulled into a terminal, or continue from a phone through Remote Control, which suits long tasks you check on between meetings.
Claude Code plans and pricing
| Plan | Price | What it includes |
|---|---|---|
| Pro | $20 per month, $17 billed annually | The first individual plan that lists Claude Code (Free does not). Adds more usage than Free, Opus access and the full set of Claude apps; usage limits apply. |
| Max | From $100 per month | Choice of 5x or 20x the usage of Pro, higher output limits, priority access at busy times and $100 to $200 per month in API credits for building on the Claude Platform. |
| Team standard seat | $20 per seat per month billed annually, $25 monthly | For teams of 2 to 150. Includes Claude Code, more usage than Pro, central billing, single sign-on and usage analytics, with no model training on content by default. |
| Team premium seat | $100 per seat per month billed annually, $125 monthly | Five times the usage of a standard seat for heavy Claude Code users. Seat types can be mixed inside one Team plan. |
| Enterprise | $20 per seat per month plus usage at API rates | Quoted on the pricing page as billed annually with a 20 seat minimum. Adds audit logs, SCIM, role-based access, admin spend limits and custom data retention controls. |
Prices as of October 2026; vendors change plans often, so confirm on the vendor's pricing page.
What Claude Code does
- CLAUDE.md and auto memory
- A CLAUDE.md file in the project root is read at the start of every session and holds coding standards and review checklists. A repository that already has an AGENTS.md can be read in its place. Claude also saves learnings between sessions on its own.
- Skills and hooks
- Skills package repeatable workflows such as a pull request review or a staging deploy so a team can share them. Hooks run shell commands before or after actions, for example formatting after every file edit.
- Parallel and background agents
- A lead agent can spawn other Claude Code agents for different parts of a task and merge the results. Background agents run several full sessions that you watch from one screen.
- Routines and scheduled tasks
- Recurring work such as morning pull request reviews or weekly dependency audits can run as routines in the cloud, even with the laptop closed, or as desktop tasks on your own machine.
- Agent SDK
- The Agent SDK exposes Claude Code's tools and permissions so you can build your own agents with full control over orchestration, which matters if the terminal workflow is only the starting point.
- MCP connections
- Model Context Protocol servers let Claude Code read design docs in Google Drive, update tickets in Jira or pull data from Slack, using servers you configure yourself.
How Claude Code handles your data
Anthropic's data usage page separates consumer from commercial accounts. For Free, Pro and Max, Anthropic trains new models on your data, including Claude Code sessions, only when the model-improvement setting is on. Team, Enterprise, API and third-party platform use is not used for training under commercial terms unless the customer joins the Development Partner Program. Retention is 5 years if you allow training and 30 days if you do not, and 30 days on commercial accounts. Zero data retention is available to qualified Enterprise accounts on request. Transcripts are also cached locally for 30 days by default. Transcripts sent through /feedback are kept for 5 years.
Where Claude Code falls short in practice
- Long sessions hit usage windows. Subscription plans state that usage limits apply, and the pricing page frames Max as 5x or 20x Pro. A heavy agent day on Pro is the moment the allowance runs out, and the fix is a bigger plan or API billing.
- The terminal is the center of gravity. JetBrains support is a plugin that needs the command line tool installed separately. Developers who want suggestions to appear as they type will find the product built around delegating tasks instead.
- Seat price hides the real bill. Enterprise is listed as $20 per seat plus usage at API rates, and the pricing page says usage cost scales with model and task. Forecasting a team budget needs a pilot, not a calculator.
- Local transcripts sit unencrypted. Session transcripts are stored in plaintext under the user's home directory for 30 days by default. Shared or managed laptops need a policy for that folder, or a shorter cleanup period.
Works with
Alternatives to Claude Code
Claude Code: common questions
Do I need a paid plan to use Claude Code?
Yes for the subscription route. Anthropic's pricing page marks Claude Code as not included on the Free plan and included on Pro, Max, Team and Enterprise. The overview also says most surfaces need either a Claude subscription or an Anthropic Console account, so API credits are the other way in.
Does Claude Code train on my code?
It depends on the account. On Free, Pro and Max, Anthropic trains on data, Claude Code sessions included, when the model-improvement setting is on, and you can change it at any time. On Team, Enterprise and API accounts, Anthropic says it does not train on code or prompts sent under commercial terms.
How long does Anthropic keep my Claude Code data?
On consumer plans the period is 5 years if you allow model improvement and 30 days if you do not. Commercial accounts get a standard 30 days. Zero data retention exists for qualified Enterprise accounts and is switched on per organization after eligibility is confirmed.
Can Claude Code run inside my existing IDE?
Yes. Anthropic documents a VS Code extension with inline diffs and a JetBrains plugin for IntelliJ IDEA, PyCharm and WebStorm. The JetBrains plugin requires the command line tool to be installed separately, so the IDE is a front end for the same engine.
Can Claude Code be used with a model provider other than Anthropic's own API?
Anthropic states that the terminal, VS Code and JetBrains versions support third-party providers, and its data page names Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry among them. Some features differ by provider, so check the feature availability page before committing.