Permission choices before a coding session
Claude Code takes a programming task and codebase context, then works through the task with permission controls that determine when a person must approve actions.
Claude Code is Anthropic's coding agent for working directly with a codebase through the terminal and connected development surfaces. A programming task and code context go in; the expected output is proposed or applied engineering work, with permission controls deciding when actions need a person's approval. It is a way to keep an agent inside a deliberate development workflow, not a substitute for ownership of the code, tests, or release.
Job
Choose Claude Code when a developer wants help exploring, changing, debugging, or shipping work in an existing codebase while retaining a deliberate level of approval. The ordinary job is broader than inline completion: it can work from a terminal and alongside development tools while a person decides how much interruption and oversight a session needs.
Input → output
The practical input is a coding objective, relevant codebase context, and a chosen permission posture. Claude Code can read context and request or perform work according to that posture; the user receives its progress, proposed changes or actions, and approval prompts where the selected mode requires them. The exact actions, tool integrations, active model, and result quality depend on the environment, project, account, and prompt.
Workflow shape
- Choose the current Claude Code surface and a permission posture that matches the task's sensitivity and the team's review expectations.
- State durable technical boundaries through the documented permission controls rather than relying on a conversational request alone.
- Review the actions, code changes, and test results that matter to the task before accepting a change.
- Keep version control, CI, secret handling, environment isolation, and release approval as separate team controls.
The tool's permission settings shape the approval rhythm of a session. They do not transfer responsibility for a repository or make a more permissive session safe by themselves.
Choice boundary
Claude Code fits supervised development work where an accountable person can choose the level of approval and verify the resulting changes. It is not the right choice when an organization needs a confirmed account-specific permission state before it can begin, when external access or credentials cannot be scoped, or when unattended production changes need to be accepted without independent review. Permission controls can reduce prompts; they are not a security, compliance, compatibility, or deployment approval.
Access and price
The current product page lists Claude Code for macOS, Linux, and Windows, with terminal, web, mobile, IDE, GitHub, and Slack surfaces. It lists these access routes: Claude Pro, Claude Max, a Team or Enterprise premium seat, or a Claude Console account. The documentation also makes availability of some permission modes contingent on account, organization, model, provider, and managed-policy conditions. No sign-in, organization policy, payment flow, or account-specific mode was used for this tool record.
| Route | Current English pricing-page price and cadence | Claude Code and usage boundary |
|---|---|---|
| Free | $0 | The product page does not list Free as a Claude Code access route; do not assume entitlement from the Free price alone. |
| Pro | $17/month with annual subscription ($200 billed up front), or $20/month billed monthly | Claude Code is included. Usage is not a fixed task or message quota: it varies by model, task, and feature use, with rolling five-hour and paid-plan weekly limits. |
| Max 5x | $100/month | Claude Code is included; the plan is described as five times Pro usage. Usage limits still apply. |
| Max 20x | $200/month | Claude Code is included; the plan is described as 20 times Pro usage. Usage limits still apply. |
| Team Standard (2–150 people) | $20/seat/month billed annually, or $25/seat/month billed monthly | The current product page specifically names a premium Team seat as a Claude Code route. Do not infer Standard-seat Claude Code access without checking the account's current entitlement. |
| Team Premium | $100/seat/month billed annually, or $125/seat/month billed monthly | The product page lists a Team premium seat as a Claude Code access route; the pricing page describes five times the Standard-seat usage. |
| Enterprise | $20/seat/month plus usage at API rates; annual billing | The product page names an Enterprise premium seat as a Claude Code route. A separate public premium-seat uplift or negotiated contract price was not shown on the checked pages, so that part is unknown. |
| Claude Console | Standard API pricing, metered by model and tokens | Claude Code consumes API tokens; the current product page does not establish a fixed all-in coding-session price or a Code-specific token allowance. |
For the Console route, the current price page lists the following standard API rates per million tokens. These are model rates, not a promise that a particular Claude Code session will use that model or consume a predictable number of tokens.
| Current model listing | Input | Output | Prompt-cache write / read |
|---|---|---|---|
| Fable 5 | $10 | $50 | $12.50 / $1 |
| Opus 5 | $5 | $25 | $6.25 / $0.50 |
| Sonnet 5 | $2 introductory / $3 standard after 2026-08-31 | $10 introductory / $15 standard after 2026-08-31 | $2.50 / $0.20 during the listed introductory period |
| Haiku 4.5 | $1 | $5 | $1.25 / $0.10 |
The price page says displayed prices exclude applicable tax and can change. It also lists 1.1× input/output pricing for US-only inference and 2× standard pricing for Opus 5 Fast Mode. The product page describes Fast Mode as a higher-cost Claude Code research-preview configuration, but this handoff did not verify availability for any account. Regional currency, tax, free trials or credits for Claude Code, the current model selected by a subscription session, and any additional usage-credit charge are unknown here and must be checked at the point of purchase or use.
Verification state
The official Claude Code product page, permission-mode documentation, and pricing page were re-read on 2026-08-08. This page records documented product structure and current English-page prices; it does not reuse the linked article's local fixture, command sequence, settings checkpoint, screenshot explanation, recovery chain, or conclusion.
The catalog mapping is /tool-logos/claude-code.webp. Its official source is the inline aria-label="Claude" wordmark on the current Claude Code product page, a native SVG with a 0 0 573 125 viewBox. The page did not expose a separate standalone Claude Code mark. Local candidate SHA-256 is 19d1a4eedb2cc6f70077b0a3434bf010bd502205fe3f8f990fedf7c619be3e23; the fixed media publisher converts the exact SVG to WebP and performs public readback. It must not be replaced by a crop or unofficial recreation.