Claude Code
Anthropic AI agent: give it a task in words, then it works on its own.
Sign in with $17/mo
Pro with annual billing
Claude models
Anthropic models only
Not directly from Russia
Requires a foreign IP address and payment method
Closed source
Repository - tracker only
Use it for serious work on your project. Do not use it for editor suggestions or free access: there are no suggestions here, and no free plan either.
Claude Code is not an editor or a chat, but an autonomous executor. You describe a task in words, then the agent works on its own: finds the right place in the project, makes changes, checks its work, and brings back a finished result with a report on what changed. It feels like a contractor you assign a task to, not a tool you operate by hand. It was born in the terminal, but by August 2026 its main entry point is its own desktop app.
1. Should you get it
The choice comes down to two profiles: see which one is yours.
Who it is for. Claude Code is built for work on a live project - at work or your own. Redo half a site, clean up tangled code, fix what broke: the agent finds the right place, plans, and checks itself. It handles long tasks without supervision, reporting along the way.
Don't write code? The terminal is optional: you can assign a task in the desktop app or directly in the browser, like in a chat. But the product is still about code - to accept the agent's work, you need to understand what it did.
Who it is not for. There are no suggestions while you type code - use Copilot and Cursor for that. There is no free Claude Code: Anthropic only offers the chat for free, while the agent starts at $17/mo. Claude models only; you cannot connect your own local ones. It does not officially work from Russia; details in section 5.
Just getting started? Haven't written code or built projects yet - start with the page“What is vibecoding”, then come back here for a tool.
It fits - use it in section 3. If you are choosing between it and Codex, the fork is in section 7.
Claude Code news
- Claude Code Projects work on your computer: local support addedSeptember 24
- Claude Code keeps working after you close your laptop: Cloud sessions leave previewSeptember 24
2. What it does
The easiest way to show it is in practice: Claude Code builds this site too. Below are six real tasks it completed here over the past month.
Rewrote the page in a day. The Codex page was a month old, and its facts were outdated. The agent reread 128 of our news stories about it, checked every number against fresh sources, and rewrote the page from scratch - on August 17, it was current again.
Delivered research overnight. It received a question in the evening and delivered a report by morning: about forty search queries, 45 sources read, and a conclusion added to the page. It worked on a separate laptop while people slept.
Compiled a biography and verified the facts itself. A founder biography in Russian and English: it cross-checked research papers against arXiv, and deal dates against publications from those years. Anything it could not verify, it noted as such.
Found where the money was leaking. The hosting bill was melting before our eyes: $15 out of $20 in six days. The agent checked billing, found the cause, and suggested another plan - $103 less per month.
Fixed what failed overnight. A nightly demand-measurement script started failing. The agent found the network cause, rewrote the script with retries - the failures stopped.
On duty without days off. Claude Code also runs the site's news feed: on a separate laptop, it checks every minute whether a story is ready, selects stories, and writes copy without a human.
What it looks like. The loop in any of these tasks is the same: you write what you need, the agent lays out a plan, makes edits, shows every change line by line, and sees it through. The full loop is in the frames below.
How much freedom to give it. By default, the agent asks for confirmation for every edit and command. Once you trust it, turn on auto mode: edits are accepted automatically, stopping only for risky actions.
Where it stops. The agent does not take dangerous actions even in auto mode: in one of our runs, it refused to click the permissions-granting button - a human made the final click.
Then on its own. Higher up the scale - loops, background tasks, and parallel subagents in one session. The top tier: cloud sessions on claude.ai/code work while the laptop is closed, and you can manage them from your phone.
3. Where it works and how to install
One subscription unlocks 7+ surfaces: desktop, terminal, IDE, web and iOS, CI, Slack, SDK. Start with desktop - today it is the main way to use Claude Code: parallel sessions, diffs, and tasks without the command line. Terminal is for those who live in the CLI: it has the same capabilities and the full set of flags.
Before you start. You need a Claude account with a Pro/Max subscription or a Claude API key. Installers for all surfaces are on the official Claude Code page.
You can install it from Russia. You can install it from Russia. First, switch your Russian IP address to a foreign one: American, Dutch, or any other. The service is officially unavailable in Russia, so without changing your address, the site simply will not open. Second, you will need a foreign card - working payment methods are collected in section 5, with a step-by-step breakdown. Among local options, Yandex SourceCraft is the closest fit: an IDE assistant, not an autonomous agent.
On desktop+
App for macOS and Windows, including ARM64: parallel sessions without a terminal, scheduled tasks, and phone launches through Dispatch.
Step 1. Download the app at claude.ai/download
Step 2. Sign in using the method offered by the app
In the terminal+
Native surface and the full set: auto mode, subagents, hooks, /usage command. Start here if you want maximum control.
Open Terminal
This is an app where you enter commands as text. Press ⌘ Space, type "Terminal," and press Enter.
Paste the installation command
Copy it using the button on the right, paste it into terminal (⌘ V), and press Enter. The script will download and install Claude Code automatically — it takes about a minute.
curl -fsSL https://claude.ai/install.sh | bashAlternative if you have Node.js 18+: npm install -g @anthropic-ai/claude-code
Launch it
Type claude and press Enter.
claude✻ Welcome to Claude Code!
/help — tips · try: "explain this project"
Sign in to your account
When you first launch it, Claude Code will open a browser and ask you to sign in. Have a Claude subscription (Pro or Max)? Choose it — no separate payment needed. Don't have one? Choose Anthropic Console and pay for usage. Once you've confirmed sign-in, you're ready to work.
In the editor+
VS Code extension (also installs in Cursor) and JetBrains plugin: changes are visible directly in files, and you review agent plans without leaving the editor. If you need an agent alongside the editor, not instead of it.
code --install-extension anthropic.claude-codein JetBrains: Settings, then Plugins, then "Claude Code"
In the browser and on iOS+
Cloud sessions at claude.ai/code: nothing to install, code runs in Anthropic's sandbox. You can continue a session from your phone or pull it into your terminal with the teleport command.
Open claude.ai/code; there is nothing to install
In CI+
Automatic review of every PR and issue analysis. For teams that need an agent in the pipeline, not the editor.
In Slack+
@Claude in a thread turns discussion into a pull request. Channels push events from Telegram, Discord, and webhooks straight into a running session.
For your own agents+
Your own agents on the Claude Code engine (TypeScript and Python) and cloud schedules on Anthropic infrastructure. For those building on top, not just using it.
4. What the tests show
An agent score is always a pair: "agent + model": Codex is measured with GPT-5.6, Claude Code with Fable and Opus. Compare pairs, not brands.
Claude Code's best pair is with Claude Fable 5.1 max (with fallback). In the 23.09.2026 measurement, it scored 62.2 aa index - rank #1 out of 6 pairs in the table. The current leader is Claude Code with Claude Fable 5.1 max (with fallback): 62.2.
| # | Agent | AA score | RuBench | |
|---|---|---|---|---|
| 1 |  CodexOpenAI | 253 | 80GPT-5.6 Sol | 66,7GPT-5.5 · xhigh |
| 2 | Claude CodeAnthropic | 208 | 77Fable 5 | 78,7Opus 4.8 · max |
| 3 |  CursorAnysphere | 187 | 62GPT-5.5 | — |
| 4 |  GitHub CopilotGitHub | 30 | — | — |
| 5 |  AntigravityGoogle | 27 | — | — |
| 6 |  ReplitReplit | 26 | — | — |
| 7 |  LovableLovable | 23 | — | — |
agent table: weight - editorial · AA score - best pair · AA Coding Agent Index 23.09.2026 · Terminal-Bench 4.0 23.09.2026 · DeepSWE v1.1 22.09.2026 · “—” - pair not measured
Choosing between it and Codex - the duel in section 7.
5. How much it costs and how to pay from Russia
Entry - $17 per month. This is the Pro plan when paid annually; monthly, the same Pro costs $20. It includes full Claude Code on all surfaces. There is no free option: the Claude Free plan does not unlock the agent, and an App Store subscription does not include it either.
Plans and what they include
| Plan | Monthly price | What it gives you |
|---|---|---|
| Pro | $17 when paid annually, $20 monthly | full access to Claude Code - all surfaces and models |
| Max 5× | $100 | 5× Pro limits - daily work without hitting the ceiling |
| Max 20× | $200 | 20× limits - long agent sessions and parallel tasks |
| Team | Premium seat $100-125 | for teams; Claude Code is available only with a Premium seat - the regular $20-25 seat does not include it |
| API | By tokens | without a subscription: Fable 5 - $10/$50 per million tokens, Opus 4.8 - $5/$25, Sonnet 5 - $2/$10 |
prices checked on August 16, 2026 against Anthropic's official pricing
Limits are shared across everything. The subscription shares one pool across claude.ai, Claude Code, and desktop: 5-hour sessions plus a weekly bar. You can see usage with the /usage command. Hit the ceiling - buy extra credits at API prices (up to $2000 per day), and since August 13 desktop can continue working on its own after the limit resets.
You can pay from Russia. A Russian card will not work: Stripe declines it before the charge. There are three working routes - a foreign card (Armenia, Kazakhstan, Kyrgyzstan; the author's main method), an intermediary for rubles (MTS, Banki.ru, restore:), and a virtual card topped up in USDT. The $20 face value costs 1600-3200 ₽ per month; the step-by-step breakdown is behind the door below.
What we checked and how
| Fact | What we know | Verified |
|---|---|---|
| Plans | No free plan. Entry point - Pro: $17 per month with annual billing, $20 monthly. Max adds limits: 5x for $100, 20x for $200. Teams need Team, and there is a catch: Claude Code is included only in a $100–125 Premium seat; a regular $20–25 seat does not include it. Enterprise is priced differently: $20 per seat plus usage at API rates. No subscription is needed through the API - pay only for tokens. | 2026-08-16 |
| Free access | No. The free Claude plan provides chat only; the agent is not included. The cheapest entry point is Pro at $17–20 per month or the API with token-based billing. | 2026-08-16 |
| Add-ons beyond limits | Hit a limit? Buy credits at the same price as the API. Top-ups trigger automatically at a set threshold, with a maximum spend of $2000 per day. Credits add capacity but do not extend the 5-hour session cycle. | 2026-08-16 |
| Limits | Limits are twofold: a 5-hour session plus a weekly cap, both overall and per model. There is one pool for chat, Claude Code, and desktop: spend it in chat, and it is spent for the agent too. Usage is shown at claude.ai/settings/usage and with the /usage command. One nuance: Anthropic’s help center describes which model the weekly cap applies to inconsistently, so until official clarification, assume both an overall and per-model cap. | 2026-08-16 |
| Model pricing (API) | API pricing per million tokens, input and output: Fable 5 - $10 and $50; Opus 4.8 - $5 and $25; Sonnet 5 - $2 and $10, and its introductory price became permanent: the September 1 increase was canceled. Sonnet 4.6 - $3 and $15; Haiku 4.5 - $1 and $5. | 2026-08-16 |
| For students | There is no free student plan. A university can adopt Claude for Education, a shared plan with Claude Code, with pricing on request. Individual students have the Campus program, which provides API credits. | 2026-06-10 |
All vendors rebuilt billing around token credits in 2026; figures change every few weeks. Each line above has a verification date: do not trust figures without a date, including ours - and check them against Anthropic's official pricing.
Working with AI agents? Here is what it looks like over the long haul63 days, 4,642 commits, not a line by hand. A live feed of this site being built.Watch live
6. Can you trust it with code
What to check before letting an agent into a production repo.
Where the code goes. Under commercial terms (Team, Enterprise, API), Anthropic does not train models on your code. On personal Pro and Max plans, training happens only with consent: data is retained for 5 years when enabled, 30 days when disabled.
What the agent can break. It edits files and runs commands. Since August 14, new sessions run in auto mode: a separate classifier filters out dangerous commands. You set the boundaries yourself - down to confirming every step; the enterprise plan adds organization-wide controls.
What we checked and how
| Fact | What we know | Verified |
|---|---|---|
| Code privacy | Under commercial terms (Team, Enterprise, API), Anthropic does not train models on code and prompts. Private plans (Free/Pro/Max): training occurs only with user consent, with data retained for 5 years when enabled and 30 days when disabled. Enterprise can enable zero data retention (ZDR) for the organization. | 2026-06-10 |
| Organization controls | Admins centrally set policies: the allowed-model list and its enforcement, allowed and blocked MCP servers, minimum and maximum CLI version, mandatory organization sign-in, and monitoring through OTel. Since v2.1.200, model overrides for the managed list are taken only from organization settings. | 2026-07-04 |
| Availability from Russia | Russia is not on Anthropic’s official list of supported countries. Installation docs require being in a supported country: it is a system requirement, like the Node version. | 2026-06-10 |
| Open source | No: the CLI is proprietary, and the license refers to Anthropic's commercial terms. The github.com/anthropics/claude-code repository contains only the changelog, issues, examples, and plugins. No source code. | 2026-06-10 |
7. Claude Code or Codex
The fork is not “who is better,” but “what kind of work.” Claude Code wins on depth: one agent can carry a complex task in a large repo for a long time and make careful edits. Codex wins on breadth and price: it is an assembly line run in batches.
In August, they diverged even further. Samsung brought Claude Code into chip development: verification shrank from a month to two days (August 13). Over the same period, Codex moved beyond code - it clicks through your apps, fixes Windows settings, and creates thumbnails for YouTube.
What to choose for the task
| Criterion | Claude Code | Codex |
|---|---|---|
| Strength | depth: long complex tasks in a large repo, careful edits | breadth: task pipeline in batches, and not just code |
| Fresh proof | Samsung: chip verification in 2 days instead of a month (13.08) | IBM moves 150,000 consultants to development via Codex (13.08) |
| Entry and price | from $17/mo, no free plan | free entry through ChatGPT, paid plans raise limits |
| Hosted in | 7+ surfaces: terminal, IDE, desktop, web and iOS, CI, Slack, SDK | 7 surfaces: ChatGPT app (macOS, Windows, Linux from 11.08), CLI, IDE, cloud, GitHub, phone, SDK |
| Beyond code | focus on repo and terminal | clicks and types in your programs: fixes Windows, makes covers |
| From Russia | officially unavailable | officially unavailable |
Live install count. For a developer, npm works like an app store: terminal agents are installed from there, and every installation appears in a public counter. The chart adds up installs for the week - the higher the line, the more often it is being installed now.
Install race: Claude Code vs Codex
npm downloads per day · ● live
Pros keep both. Gergely Orosz, author of The Pragmatic Engineer newsletter, gave first place in the terminal to both Codex and Claude Code on August 6. They split tasks by type - complex work goes to Claude, bulk work to Codex.
8. History
A year and a half from a terminal experiment to the most popular coding agent - with the drama of Fable 5 disappearing and returning in the middle.
24.02
Born as a research preview with Claude 3.7 Sonnet: an agent living right in the terminal; access via a waitlist.
22.05
Leaves preview: general availability with the Claude 4 launch — Pro and Max subscriptions open the agent to everyone.
29.09
Grows up to version 2.0 on Sonnet 4.5: checkpoints, hooks, and SDK becomes Agent SDK.
20.10
Breaks out of the terminal: web version and iOS preview — agent sessions from the browser, environment sandboxing.
24.11
Gets the Opus 4.5 engine — Anthropic calls it the world's best model for code and agents.
14.04
Desktop rebuilt for parallel agents: multisessions, panels, Routines.
01.07
Fable 5 returns to full availability: export controls lifted, redeployed with a new security classifier.
9. Frequently asked questions
What is Claude Code in simple terms?+
An AI agent for coding that lives in the terminal. Give it a task in plain text, and it edits project files, runs checks, and takes the work through to a commit. Made by Anthropic, the same company behind Claude models.
Claude Code or Codex: which should you choose?+
Claude Code is stronger in agentic environments, design/frontend tasks, and large parallel runs; Codex practitioners praise its thoroughness and harness quality. The working pattern as of June 2026 is to use both for different task types.
Claude Code or Cursor: which should you choose?+
They are different categories. Cursor is an AI editor: you write code yourself, while AI helps with autocomplete and in-place edits. Claude Code is an autonomous agent: you give it the whole task, and it plans, edits files, and takes it through to a commit, including in the background and in parallel. Live in the editor - Cursor; delegate tasks - Claude Code. Many pros keep both: Cursor for quick hands-on edits, Claude Code for larger chunks of work.
How much does Claude Code cost?+
The Free plan does not include Claude Code. Pro - $17/month billed annually ($20 monthly), Max 5x - $100/month, Max 20x - $200/month, Team - from $20/seat, Enterprise - $20/seat + usage at API rates. Above the limits, usage credits are available at standard API rates (checked 06/10/2026).
Does Claude Code work in Russia?+
Russia is not on Anthropic's official list of supported countries, and the installation docs require you to be in a supported country. This is the official position as of 06/10/2026; this card provides no workarounds - practical access scenarios from Russia are covered in separate guides. Among local options, the closest is Yandex SourceCraft Code Assistant (agent mode, payment in rubles), but it is not a direct equivalent of a terminal-first agent.
Does Claude Code have a free tier or student access?+
There is no free tier: the Claude.ai Free plan does not include Claude Code. There is no public free student plan either - there is Claude for Education (a university-wide plan with Claude Code, pricing on request) and the Campus program with API credits (checked 06/10/2026).
Do you need a Claude Code API key?+
No, if you have a Claude Pro/Max subscription: sign in through the browser, no keys required. An API key is the second path: token-based metering instead of a subscription; it is needed for CI, automations, and team pipelines without a personal login. One of the two paths is required - there is no free sign-in.
What is Claude Code's context window?+
The standard window is 200K tokens where extended context is not enabled. The 1M window is available with Fable 5, Sonnet 5, Opus 4.6+, and Sonnet 4.6: Sonnet 5 on the Anthropic API is always 1M; for Opus it is automatic on Max/Team/Enterprise, while Pro and Sonnet 4.6 use usage credits. There is no per-token price premium above 200K.
Why does Claude Code use limits quickly, and how can you save?+
Limits are shared: a 5-hour window plus a weekly bar, with one pool for claude.ai, Claude Code, and Desktop - so "Claude ate everything at once" usually means a heavy mode: Opus, ultracode, parallel subagents. Track it with /usage in the CLI (breakdown by skills, subagents, plugins, MCP) and claude.ai/settings/usage. To save: use a cheaper model for simple tasks, provide a precise context slice instead of the whole repo, and batch heavy work into background sessions. A breakdown of tactics is in the token-maximization guide.
Claude Code or SourceCraft Code Assistant (Yandex)?+
SourceCraft Code Assistant is the closest Russian option: it works from Russia with Russian payment methods, but it is an IDE assistant, not an autonomous terminal agent - it lacks background sessions, subagents, and Claude Code-level cloud mode. Need a tool strictly from Russia without a foreign card - SourceCraft is a realistic compromise; need full autonomy - Claude Code with foreign payment.
Which is correct: Claude Code or Klod Kod?+
The official name is Claude Code (Anthropic does not localize it). Russian-language texts may use "Klod Kod" and "Klaud Kod" - transliterations of the same product. For documentation and search, the original name Claude Code is more reliable.
How long do Claude Code limits really last?+
There are no official hourly figures: limits are measured in tokens and depend on the model and effort level (5-hour sessions + weekly bars, one shared pool with claude.ai and Desktop). Signals from practitioners in spring 2026: heavy agentic tasks on Pro hit the limit quickly, Max 5x/20x support a full workday, while parallel subagents and dynamic workflows burn usage many times faster than regular chat. These are observations, not official data - monitor usage through /usage.
What we wrote about Claude Code
Useful reading before you start
- What is vibecoding: starting from scratch.
- Loops: how to give an agent recurring work.
- The big Claude Code guide - 2026.
- MCP: how an agent connects to your services.
- How to pay for AI services from Russia.
- The full AI agent ranking.