Claude Code Platforms and Interfaces

If you've only ever run claude in a terminal, you've met one face of a tool that has at least seven. The same underlying engine — the agentic loop, the built-in tools, the permission system covered in Claude Code Fundamentals — also runs inside VS Code and JetBrains IDEs, inside a dedicated desktop app, in the cloud at claude.ai/code, on your phone, from a Slack channel, and even driving a browser or your literal mouse cursor. None of these are separate products bolted onto Claude Code; they're different windows onto the same session model, and in most cases they read the same CLAUDE.md, the same .claude/settings.json, and the same MCP server configuration.

That shared foundation is exactly why surface choice is a real decision and not just aesthetic preference. Picking the wrong surface for a task costs you real things: a long migration started in a terminal session dies the moment you close your laptop lid, while the same task started as a cloud session keeps running. A UI bug investigated by reading code alone takes three times as long as one investigated with a screenshot and a live browser. A permission mode that's perfectly safe on your own machine is actively dangerous run against a business-critical service from someone else's device. This subject is about building the judgment to make that call quickly, plus the mechanical facts — what each surface actually supports, what it silently drops, and where the docs contradict a surface's marketing — that the call depends on.