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.