What's the difference between the Claude Agent SDK and Claude Code?
The Claude Agent SDK is the same agent loop, tools, and context management that runs Claude Code, packaged as a Python or TypeScript library so you build your own agent on it instead of using Anthropic’s own interactive product. Claude Code is that product: the terminal, IDE, and web surfaces where you use Claude yourself, all running the identical loop underneath.
The two diverge past the shared loop. An SDK agent doesn’t get Claude Code’s own system prompt by default, you opt into it explicitly with a preset, so out of the box it’s a different agent, not Claude Code wearing your branding. The SDK also adds one thing the interactive product has no equivalent for: an in-process MCP server, letting you hand Claude a function that runs inside your own application instead of a separate server process whose tool schema can drift out from under a cached client.