Anyone who follows a coding agent from a phone eventually asks the same question: will this put my account at risk?
Kiteline, an app for following coding agents (Claude Code, Codex and a dozen others) from an iPhone, iPad or Apple Watch, answers with three boundaries.
It does not read or store Claude account credentials, does not log in on the user's behalf, and does not call Anthropic's API directly. It only reads the session records Claude Code writes locally and uses the extension points and terminal that Claude Code exposes.
The phone talks to the user's own computer (directly on a LAN, otherwise peer-to-peer or through a relay). From Claude's side nothing changes: same machine, same network environment, no new device.
An optional feature runs Claude Code in the background on a remote machine so VS Code does not need to stay open. It is off by default and enabled per device; the exit IP is checked and shown before enabling; and the background process must use the same proxy, certificates and API endpoint as Claude Code inside VS Code. If those cannot be read, it falls back to VS Code and says why.
Keep one machine and one network exit, do not share an account, do not hand credentials to third-party tools, and use the phone as a remote control only.
Claude Code is Anthropic's product; Kiteline is an independent third-party tool. The desktop side is open source on GitHub; the phone app is iOS/iPadOS only (no Android).