Claude.ai vs Claude Code vs the API

Same model, three front doors. Picking the wrong one is the single most common reason people think Claude is "fine but slow" — they're hand-driving a task that should be automated, or scripting something that wanted a conversation.

The quick decision

• Thinking, drafting, exploring? → Claude.ai
• Doing work on real files and tools? → Claude Code
• Wiring Claude into a product or pipeline? → the API

Claude.ai — the thinking surface

The web and desktop app. Best for conversation, ideation, writing, and anything where you want to see and steer each step. Projects and Skills live here. This is where most knowledge work happens.

Claude Code — the workbench

A terminal agent that reads and edits files, runs commands, and verifies results. Best when the work touches your actual machine: a repo, a folder of documents, a data file. It can run multi-step tasks unattended and check itself.

The API — the engine

For builders. You call Claude from your own software, on your own triggers, at scale. This is how you put Claude inside a product: a support classifier, a content pipeline, an in-app assistant.

from anthropic import Anthropic
 
client = Anthropic()
msg = client.messages.create(
    model="claude-sonnet-4.5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Summarize this ticket: ..."}],
)
print(msg.content[0].text)

A simple rule

Ask: who pulls the trigger? A human, every time → Claude.ai. A human, on real files → Claude Code. Your software, automatically → the API.