Claude is available through several distinct surfaces, and picking the right one is itself a skill — not just a login choice.
| Entry point | Best fit for |
|---|---|
| claude.ai (web / mobile / desktop) | Conversational work, one-off tasks, most knowledge-worker use cases |
| Claude Code | Writing, debugging, or shipping code — CLI, desktop, or IDE-embedded |
| Claude Cowork | Heavier multi-step work: research, analysis, long-form writing, tasks spanning many tools/files |
| API / Console | Building Claude into your own product or automated pipeline |
| In-app integrations (Chrome, Excel, PowerPoint) | Working directly inside a browser, spreadsheet, or slide deck without switching tools |
| Claude Tag (Slack) | Team workflows — anyone can tag @Claude into a channel thread |
Decision criteria:
- Who's doing the task? A developer shipping code has different needs than an ops manager triaging email.
- What shape is the task? Purely conversational vs. agentic (many steps, tool calls) vs. embedded in an existing document/app.
- Does it need to touch other files or apps? If yes, an integration or Cowork often beats a plain chat window.
Common misconception to unlearn: assuming the API and claude.ai always expose identical capabilities. They don't — some features are chat-surface-specific (e.g., Artifacts rendering) while others are API-specific (e.g., raw tool-use control). Matching entry point to task means knowing which surface actually has the feature you need.