Claude Seekho

Big Picture 1: Where a Request Comes From

Claude is one model family with several entry points. Whichever one you use, the same thing happens underneath: a Messages API request is built and sent to the same model. This stage is the map of those entry points.

From surface to Messages API request
  1. 1.You ask Claude a question on some surface
  2. 2.Surface: claude.ai apps, Claude Code, Claude API + SDKs, Agent SDK, Managed Agents or Cloud
  3. 3.The surface builds a Messages API request
  4. 4.The same model family answers it
  5. 5.Next stage: what is inside this request

Every surface ends in the same kind of request.

The entry points, one by one

  • claude.ai / Desktop / Mobile: the most familiar entry point. You chat, create Projects, see Artifacts, run Research and can even talk by voice.
  • Claude Code: built for developers. It runs as a CLI, inside VS Code and JetBrains, on the web, and in Slack and GitHub Actions. Claude works directly on your code.
  • Claude API + SDKs: your program sends a POST request to /v1/messages. SDKs exist for Python, C#, TypeScript, Java, Go and more, so you do not write raw HTTP.
  • Agent SDK: the same loop that powers Claude Code, running inside your own app.
  • Managed Agents: Anthropic hosts both the agent loop and the sandbox for you.
  • Cloud: if your company already lives on a cloud, Claude is available on Bedrock, Vertex AI and Foundry. Same model family, the entry point just belongs to your cloud.
  • Messages API request: the one thing every surface builds internally. Chat app or Claude Code, the same kind of request goes out and the same model answers it.