Unclear Whether anything besides both input and output being a terminal decides if the prompt can be shown.
Remote Control lets you drive Claude Code sessions in a folder from your other devices. Before, starting it with claude rc in a folder you had not yet trusted printed "Workspace not trusted" and quit. When Claude Code is running in an interactive terminal, it now shows a safety notice and asks Trust <folder>? [y/N].
The notice lists what trusting the folder turns on for sessions started there:
- credential helpers and
headersHelper - environment variables set by the project
- hooks, which are commands that run automatically at certain points
- the servers in the folder's
.mcp.jsonthat would be started - pre-approved permission rules
- extra directories added to the session
autoMemoryDirectory
If you answer yes, the choice is saved and Remote Control carries on. In your home directory, trust lasts for that run only and is not saved. If the choice cannot be saved to your config, you are told you will be asked again next time. If Claude Code cannot prompt, or the terminal is too small to show all the warnings, it still exits with the "Workspace not trusted" error rather than showing a cut-off prompt.
You no longer have to run claude separately in a new folder before Remote Control will start there. The prompt also spells out the risky project settings before you accept them.
Whether anything besides both input and output being a terminal decides if the prompt can be shown.