Feat: Hermes PC-Zugriff via MCP (executor + mcp_pc)

- mcp/mcp_pc.py: MCP-Server der PC-Tools an den Executor proxied
  (pc_shell, pc_screenshot, pc_type, pc_key, pc_open, pc_status)
- client/hermes-pc/executor.py: FastAPI auf Port 7777 (Daemon-Reg entfernt)
- client/hermes-pc/start.bat, install.bat, requirements.txt
- client/hermes-voice/agent_client.py, main.py: agent_client integration

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

client/hermes-voice/main.py

@@ -12,7 +12,7 @@ import config_manager as cfg
 from audio_input import start_recording, stop_recording, transcribe
 from audio_output import speak, stop_speaking, play_ding
 from screen_capture import capture_screen
-from ai_client import ask
+from agent_client import ask_agent, get_agent_status
 from hotkey_listener import HotkeyListener, KeyCapturer, str_to_display
 
 ctk.set_appearance_mode("dark")
@@ -145,7 +145,12 @@ class HermesApp(ctk.CTk):
         ui_event("log_user", text)
         screenshot = (capture_screen()
                       if self.settings["screenshot_enabled"] else None)
-        response = ask(text, screenshot, self.settings)
+
+        ui_event("status", ("thinking", "💭 Denke…", "#3b82f6"))
+        response, tool_calls = ask_agent(text, screenshot, self.settings)
+
+        for tc in tool_calls:
+            ui_event("log_tool", tc)
 
         ui_event("log_hermes", response)
         ui_event("status", ("speaking", "🔊 Spricht…", "#a855f7"))