[FEATURE] Verifiable human approval for external tools in Remote Control

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

I often use Remote Control to supervise coding tasks on my computer from my phone. Most of that workflow works well remotely, including ordinary tool permissions.

The gap comes when an external tool has its own rules about which actions require a decision from the human user. For example, a local tool might require explicit approval before signing a release, changing a protected setting, or carrying out another sensitive operation.

When I'm at my computer, I can authorize those actions directly. When I'm away, the external tool has no supported way to verify that I personally approved the specific action. A chat message saying "approved," or permission for Claude to call the tool, is not the same thing.

Proposed Solution

I'd like an optional, trusted approval flow in Remote Control that external tools (including MCP servers) could request.

Ideally, I could:

• Review the exact action in a trusted prompt on my phone, rather than a message composed by the agent.
• Approve or reject it explicitly, with device verification when appropriate.
• Have the requesting tool receive independently verifiable confirmation that I authorized that action, not just an assertion from the agent.

The confirmation should be bound to the requested action, so the agent cannot invent it or reuse it for a different operation.

Existing tool permissions and service integrations already work for many situations; I'm not asking to replace them or add prompts to routine work. This would help tools that enforce their own human-authorization rules while users supervise their agents remotely.

I'm open to whichever implementation best fits Claude Code's trust model.

Alternative Solutions

I can return to my computer or use a separate approval interface, but both interrupt the remote workflow. A conversational confirmation is more convenient, but it isn't sufficient when the external tool must verify authorization independently.

Priority

High - Significant impact on productivity

Feature Category

Developer tools/SDK

Use Case Example

1. Claude Code is working on my computer while I'm away.
2. An external tool reaches a protected operation that requires my authorization.
3. Remote Control shows the exact requested action in a trusted prompt.
4. I approve or reject it on my phone.
5. The external tool verifies the decision, and work can continue without me returning to my computer.

Additional Context

Related: #94345 discusses distinguishing genuine user input from agent-relayed messages; #95253 discusses host-side attestation for privileged actions.

This request is specifically about an external tool verifying a human decision on an exact operation, rather than relying on the agent to report that decision.