A Claude Code workflow with NotchDock Agent Control
Claude Code is one of three integrations with native approve/deny. Agent Control also supports Claude Code question choices and plan review with feedback, while the agent continues to run in its own terminal session.
Connect and monitor
Choose Claude Code in Agent Control’s monitored connections. Session cards show provider status and activity; open the session to return to the terminal. Keep Claude Code’s own prompts and terminal visible when a decision needs more context.
Review a permission request
When Claude Code publishes a supported permission request, inspect the displayed action and context before choosing an available response. Use deny with a steering note when the request should change. An approval applies through Claude Code’s supported hook transport; the tray does not grant a general shell permission.
Answer a question or review a plan
For supported multi-choice questions, select the answer in the request card and submit it through Claude Code’s contract. For a structured plan, read the formatted preview and either approve or send feedback. The delivered state means the bridge accepted the answer; it does not prove that Claude resumed or completed the task.
When the request is not actionable
If the target cannot be verified or a request uses an unsupported response path, use Open Session and handle it in Claude Code. Keep the request pending until the provider resolves it. Do not assume an observation-only card is an approval.
Permission checklist
Opening a terminal session can require macOS Automation permission. Optional terminal menu interaction requires explicit terminal-control opt-in and Accessibility access. The app reports targeting or permission failures instead of silently sending input.
See the Agent Control capability table and the provider overview.