Claude Cowork vs Claude Code: What's the Difference?
Claude Cowork and Claude Code run on the same agent engine but serve different work. A comparison of who each is for, where it runs, what it can reach on your computer, permissions, scheduling and plans, with examples of when to use which.
Claude Cowork and Claude Code come from the same place. Anthropic describes Cowork as using "the same agentic architecture that powers Claude Code, with no terminal required." Both plan a task, work through it step by step across tools, check their own output and hand back a result.
The difference is the work they are pointed at. Claude Code works on a codebase. Cowork works on everything else: documents, spreadsheets, research, files on your computer and tasks in your browser. This article compares the two using Anthropic's documentation, and explains which one to reach for.
Key takeaways
- Claude Code is for software: it reads a codebase, edits files, runs commands and opens pull requests, from the terminal, an editor, the desktop app or the web.
- Cowork is for knowledge work: it produces documents, spreadsheets with working formulas and presentations, organizes files and works in your browser, with no terminal.
- Cowork runs its sessions in the cloud and reaches your local files, browser and apps through the Claude Desktop app. Claude Code works directly in your project, on your computer or in a cloud session.
- Both need a paid plan. Anthropic is now merging chat and Cowork into one Claude experience on Pro and Max.
The short answer
Use Claude Code when the result is a change to software: a feature, a bug fix, a migration, a pull request.
Use Cowork when the result is a finished deliverable or an action: a report from a folder of notes, an expense spreadsheet from receipts, a slide deck from a transcript, a form filled in on a website.
What Claude Code is
Anthropic's documentation calls Claude Code "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools." It runs in the terminal, in VS Code and JetBrains, in the desktop app and on the web, and every surface uses the same engine, so a repository's instruction files, settings and MCP servers work in all of them.
Typical Claude Code work includes building features and fixing bugs, creating commits and pull requests, reviewing code automatically on every pull request, and running agents in parallel. It also runs on a schedule, in continuous integration through GitHub Actions or GitLab, and from Slack.
What Claude Cowork is
Cowork "brings Claude Code's agentic capabilities to knowledge work beyond coding." Instead of answering one message at a time, Claude takes a described outcome and works until it is done: "describe an outcome, step away, and come back to finished work."
Anthropic lists these capabilities:
- Professional outputs: "Excel spreadsheets with working formulas, PowerPoint presentations, and formatted documents."
- Direct local file access: on desktop, Claude reads and writes files in folders you connect, without uploads or downloads.
- Browser actions: Claude can open sites, read pages, click, type and fill in forms.
- Sub-agents: complex work is split into parallel workstreams.
- Scheduled tasks: tasks run on demand or on a schedule, in the cloud, "with no device online."
- Projects: related tasks grouped with their own files, instructions and memory.
Side-by-side comparison
| Claude Code | Claude Cowork | |
|---|---|---|
| Built for | Software development | Knowledge work beyond coding |
| Typical result | Code changes, commits and pull requests | Documents, spreadsheets, presentations, organized files, completed web tasks |
| Where you use it | Terminal, VS Code, JetBrains, desktop app, web | Claude Desktop for macOS, Windows and Linux (beta), web, mobile, Chrome side panel |
| Where the work runs | In your project on your computer, or in a cloud session | In an isolated environment on Anthropic's servers |
| Reaching your computer | Works in the project directory you start it in | Local files, browser and desktop apps through the Claude Desktop app |
| Approval controls | Permission modes, from asking before most actions to an auto mode reviewed by a classifier | Manual, Auto and Skip modes; explicit permission before permanently deleting files |
| Instructions | CLAUDE.md files and auto memory | Global instructions and folder instructions |
| Scheduling | Cloud or desktop scheduled tasks | Scheduled tasks that run in the cloud |
| Plans | Pro and Max individual plans, or an Anthropic Console account | Pro, Max, Team and Enterprise |
How each one handles safety
Claude Code uses permission modes. In manual mode it asks before most actions that edit files, run commands or reach the network. In auto mode, "a second model, the classifier, reviews actions instead of you." Other modes approve file edits automatically, keep Claude to planning, allow only pre-approved tools, or skip checks.
Cowork offers three modes for connector tools. Manual asks for approval. Auto has Claude review "each action for safety (such as checking for data exfiltration or prompt injection)" and block anything unsafe. Skip approves everything except blocked tools, and Anthropic advises using it only "when you completely trust every action, connector, file, app" involved. In every mode, Cowork asks before permanently deleting files, and shell commands run in an isolated environment on Anthropic's servers rather than on your computer.
When to use which: examples
| Task | Use |
|---|---|
| Fix a failing test and open a pull request | Claude Code |
| Retire a deprecated endpoint across your API, web and mobile repositories | Claude Code |
| Turn a folder of receipts into a formatted expense report | Cowork |
| Build a spreadsheet with lookups and conditional formatting from raw data | Cowork |
| Combine web research, articles and notes into a report | Cowork |
| Rename a folder of files to a consistent date format | Cowork |
| Profile slow endpoints and test optimizations | Claude Code |
| Draft a slide deck from a meeting transcript | Cowork |
Claude Cowork and chat are becoming one
On Anthropic's pricing page, the latest-news banner reads: "Claude Cowork is now just Claude. Rolling out to Pro and Max, with more plans to follow." The help center explains that the new experience combines chat and Cowork; if the message box no longer shows separate Chat and Cowork options, "every conversation can take on a Cowork task." Claude Code stays a separate product for software work.
Sharing what they make
Neither tool shares the working session itself. Anthropic lists "no session sharing" as a current Cowork limitation, though "you can share individual artifacts you create during a session." Claude Code can publish its session output as an artifact, a live page on claude.ai that you can share.
For files a team keeps editing together, such as a budget, a brief or a deck that several people and agents update, an MCP workspace such as Artifacts gives Claude Code, Codex and other agents one shared place to publish them.
Frequently asked questions
Is Claude Cowork the same as Claude Code?
No. They use the same agentic architecture, but Claude Code works on codebases and Cowork works on documents, files, research and browser tasks without a terminal.
Can Cowork write code?
Cowork runs code and shell commands in an isolated environment on Anthropic's servers to complete tasks, such as data analysis and file processing. For changes to a software project, Anthropic's tool is Claude Code.
Is Claude Cowork free?
No. Cowork is available on paid plans only: Pro, Max, Team and Enterprise.
Do I need the desktop app for Cowork?
Only for reaching things on your computer. Sessions run in the cloud and work from web and mobile, but local file access, browser use and computer use need the Claude Desktop app for macOS or Windows, open and connected.
Can Cowork keep working when my laptop is closed?
Yes. Anthropic states that "in sessions in the cloud, Claude keeps working when you close your laptop or step away," unless the task needs local files or your browser.
