Blog

Claude vs Claude Code: the chat app, the coding agent, how each is billed and which to use

Ramashankar S · · Updated · 12 min read

Claude vs Claude Code is a question of where the work happens. Claude is the chat app at claude.ai, on desktop and on mobile: you ask, it answers, writes code you copy out, and runs code in a sandbox at Anthropic. Claude Code is Anthropic's coding agent: it works inside your project, reads the codebase, edits files, runs your tests and commits, usually on your own machine.

Both run the same Claude models and come with every paid plan, and they draw from one shared set of usage limits. The Free plan includes chat but not Claude Code. This guide covers what each does, where it runs, how it's billed, and which to open for which job, checked on 10 November 2026 against Anthropic's Help Center and Claude Code 2.1.322.

Claude vs Claude Code at a glance

Claude (chat app)Claude Code
What it isA conversation: answers, drafts, explanations, ArtifactsAn agent that works in your codebase
Where you use itclaude.ai, desktop app, iOS and AndroidTerminal, VS Code, JetBrains, the desktop app's Code tab, claude.ai/code, mobile
Where the work runsAnthropic's servers; code runs in an isolated sandboxYour machine by default; a cloud VM for cloud sessions
Your codeWhat you paste, upload (30MB a file) or sync from GitHubThe whole repository it starts in
Can run your tests and gitNoYes, as your user account
Default model on paid plansYour pick in the model menuOpus 5.5
PlansFree and all paid plansPro, Max, Ultra, Team, Enterprise, or a Console API key
How it's billedPlan limitsSame plan limits, or per token with an API key
Claude chat vs Claude Code compared on purpose, surfaces, where the work runs, access to your code, plans and billing

The short version: chat produces text and files for you to use. Claude Code produces changes in your repository.

What Claude (the chat app) does

Claude, the app, is a conversation on claude.ai, the desktop app and the mobile apps. It answers questions, writes and explains code, searches the web, builds Artifacts and keeps memory across conversations. Claude's pricing page lists web search, file creation with code execution, memory, connectors and Artifacts on every plan, Free included. Opus is not on Free.

It does more coding than many people expect:

  • It can run code and create files in "an isolated, sandboxed container" at Anthropic, including Excel, Word, PowerPoint and PDF files, up to 30MB a file. File creation uses more of your limit than a plain chat.
  • The GitHub integration adds files from a repository to a chat or a project. It syncs "files (names and contents) in a repo on a specific branch" only, with no commit history or pull requests, and you press "Sync now" to refresh. The repository must fit in the context window.
  • On Pro, Max and Ultra with the combined chat and Cowork experience (rolling out since 16 September 2026), a conversation that needs files or several steps continues in a cloud sandbox without you switching modes.

What chat can't do is touch your project. It never sees files you didn't give it, can't run your test suite against your dependencies, and can't commit. You copy its code into your editor, run it, and paste errors back.

What Claude Code does

Anthropic's overview calls Claude Code "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools." You describe a task; it plans, edits across files, runs the tests and fixes failures. It stages changes, writes commit messages, opens pull requests, and can run in CI through GitHub Actions or GitLab. The current release is 2.1.322, published 10 November 2026 on npm.

You install it once and start it inside a project:

# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# then, in any project
cd your-project
claude

It asks you to log in on first run. The Claude Code install guide covers Homebrew, WinGet, VS Code and Cursor.

Claude Code runs in more places than the terminal. The desktop docs say "In the Claude Desktop app, Claude Code is the Code tab," next to chat and Cowork, and the app includes Claude Code, so you don't need the CLI. There are VS Code and JetBrains extensions, and cloud sessions at claude.ai/code or in the mobile app run in an Anthropic-managed VM against a GitHub repository and keep going after you close the laptop.

Your project shapes how it works: a CLAUDE.md file in the repository root sets coding standards it reads every session, hooks run your scripts before or after its actions, and MCP servers connect it to tools such as Jira or Slack. The Claude Code cheat sheet lists the commands.

Coding on the go with Claude CodeClaude · 4:02 · YouTube
Anthropic shows Claude Code cloud sessions: a VM per task, started from a phone, counting toward the same plan limits.

Where each one runs

This is the practical difference. Chat works on Anthropic's servers with only what you hand it. Claude Code, by default, acts on your computer.

Claude chatClaude Code (local)Claude Code (cloud session)
Runs codeIn Anthropic's sandboxOn your machine, as your userIn an isolated VM
Sees your filesUploads and synced GitHub filesThe repository it starts in, and more with your approvalA clone of your GitHub repository
Keeps going when you close the laptopNoNoYes
Uses your local tools and dependenciesNoYesOnly what the VM installs

Claude Code's security docs say the working-folder boundary is a permission prompt, so "a Bash command you approve can still write anywhere your user account can." Auto mode, where a separate classifier model reviews each action before it runs, is the default in the terminal and VS Code since v2.1.283. Turn on /sandbox or use a dev container if you want OS-level limits. Your files stay on disk, but the contents it reads go to the model like anything you paste into chat.

Cowork sits between the two: the chat app's agent for documents and spreadsheets, running in Anthropic's cloud. The Claude Cowork vs Claude Code guide covers that three-way split.

How each is billed: one plan, shared limits

Claude and Claude Code share one subscription and one pool of usage. Anthropic's Claude Code plan article says the limits "are shared across Claude and Claude Code, meaning all activity in both tools counts against the same limits." The usage-limit article names claude.ai, Claude Code and Claude Desktop as counting toward the same limit. Plans as of 10 November 2026, before tax:

PlanPriceClaude chatClaude CodeLimits
Free$0YesNoLimited, every 5 hours and every week
Pro$20 a month, or $200 a yearYesYes5-hour and weekly
Max$100 a monthYesYes5x Pro, 5-hour and weekly
Ultra$200 a monthYesYes10x Pro weekly; no 5-hour limit once verified
Team$25 or $125 a seat a month ($20 or $100 yearly)YesYesPer-seat, shared with chat and Cowork
Enterprise$20 a seat a month yearly, plus usage at API ratesYesYesSet by the admin

The pool is shared, but the two don't spend it at the same rate. Anthropic's Enterprise consumption guide rates chat "Lower intensity" and Claude Code "Higher intensity," because each Claude Code session carries system prompts, file contents, tool calls and multi-turn reasoning. The Claude Code cost docs put it more bluntly: "one debugging session can consume more than a day of chat."

How the shared plan limit is spent: a chat message sends the conversation, a Claude Code turn sends files, tool results and reasoning, and the 5-hour and weekly bars cover both

Three things catch people out:

  • A long Claude Code session in the morning can leave you rate limited in chat in the afternoon. Check Settings > Usage on claude.ai, or run /usage in Claude Code, which shows the same plan bars.
  • If ANTHROPIC_API_KEY is set in your shell, Claude Code bills that key per token instead of your plan. Run /login to switch to the subscription.
  • The monthly API allowance included with Max, Ultra and Team can't be used in Claude Code or the Claude apps.

When you hit a limit you can wait for the reset, upgrade, or turn on extra usage billed at API rates. The Claude Code pricing guide breaks down when Max or Ultra pays for itself.

Is Claude Code free?

No. Claude Code isn't on the Free plan; the cheapest way in is Pro at $20 a month ($17 a month billed yearly). The installer itself costs nothing, but it needs a paid Claude plan or a Claude Console account billed per token before it will run a task.

The only free route is a guest pass: Pro, Max and Ultra subscribers can send seven days of Pro, which includes Claude Code, to someone new to paid Claude plans. It asks for payment details and becomes a paid Pro plan unless cancelled.

Claude Code vs Claude: which to use when

Claude Code vs Claude comes down to what you want at the end. If it's an answer, a draft or a snippet, use chat. If it's a change in a repository, a passing test run or a pull request, use Claude Code.

Decision guide: which of Claude chat or Claude Code fits each kind of task, with the plan each needs
  • Ask how something works, compare libraries, explain an error message: Claude chat.
  • Write a one-off script or a regex you'll paste yourself: Claude chat.
  • Plan a feature or draft a spec before writing code: Claude chat, then hand the spec to Claude Code.
  • Fix a bug across several files, add a feature, refactor, update dependencies: Claude Code.
  • Run the tests, fix what fails, and commit: Claude Code.
  • A long task you don't want to watch: a Claude Code cloud session.
  • Documents, spreadsheets and research with no repository: chat, which hands off to Cowork on paid plans.
  • On the Free plan: chat is the only option.

The same model gives different results in the two because of what surrounds it. In chat, Claude sees what you paste and can't check its own work against your code. In Claude Code it reads the files it needs, runs the code and sees real errors. For anything larger than a snippet, that loop is the reason to switch. For quick questions, chat costs less of your limit and is faster to open.

API prices for builders

If you'd rather pay per token, Claude Code also runs on a Claude Console API key, or on Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. The chat app has no per-token option. API prices on 10 November 2026, per million tokens:

ModelInput / outputCache read
Claude Fable 5.2$8.00 / $40.00$0.20
Claude Opus 5.5 (Claude Code default)$4.00 / $20.00$0.20
Claude Sonnet 5.5$2.00 / $10.00$0.10
Claude Haiku 5.5 (prompts up to 100K)$0.10 / $0.50$0.01

Anthropic says enterprise Claude Code use averages "around $13 per developer per active day and $150-250 per developer per month." At that rate a subscription costs less for anyone coding more than a couple of days a month. The Claude Code API key guide covers per-token setup, and the token cost calculator prices your own workload.

What developers report

  • A Hacker News commenter noted that you get "vastly different outcomes from the exact same model in a different harness" (Hacker News, 28 Jun 2026, link).
  • Another commenter said people install Claude Code "to have models work inside their repositories and run the code or tests themselves" instead of copying code between an IDE and a chat (Hacker News, 13 Aug 2026, link).
  • A developer who uses Claude Code over the chat apps even for non-coding work asked, "Why use a limited version?" (Hacker News, 11 Sep 2026, link).
  • Going the other way, a commenter prefers website chatbots because they can see exactly what is sent, and pointed out that chat sandboxes can now run end-to-end tests and return a patch or zip to merge (Hacker News, 5 Nov 2026, link).
  • Anthropic's own planning guide for admins says Claude Code and Cowork "consume tokens at a significantly higher rate than standard chat" (Claude Help Center, link).

FAQ

Do Claude and Claude Code share usage limits?

Yes. On Pro, Max, Ultra and Team, chat, Cowork and Claude Code on one account all draw from the same 5-hour and weekly limits. Claude Code uses them faster per session, so a heavy coding day leaves less for chat.

Is Claude Code better than Claude for coding?

For work inside a real project, yes, because it can read your files, run your tests and fix what fails. Both use the same models, so for a short question or a self-contained snippet, chat gives an equally good answer and uses less of your limit.

Does Claude Code use the same model as the Claude app?

Yes. Claude Code defaults to Opus 5.5 on paid plans and the API, and /model switches to Sonnet 5.5, Haiku 5.5 or Fable 5.2. The chat app offers the same family in its model menu, though Opus isn't on the Free plan.

Do you need to know how to code to use Claude Code?

Not to start it, since the desktop app's Code tab needs no terminal. But it changes files and runs commands in a real project, so you need to judge its diffs and understand what it's about to run. For documents and research with no code, Cowork in the Claude app is the better fit.

Can you use Claude Code with a free Claude account?

No. The Free plan includes chat only. Claude Code needs Pro or a higher plan, a Team or Enterprise seat, or a Claude Console API key billed per token.


Sources

Checked on 10 November 2026.