Let your workflow choose the interface
Text remains the control surface; optional voice and vision features extend it when those components are installed and permissions are available.
Speech-to-text
The optional Whisper engine can convert recorded speech into text, which is then submitted as a user message.
- Install the audio component and model.
- Grant microphone access in Windows.
- Check the selected recording device.
Spoken replies
When Speak Replies is enabled, MRCO AI can read responses with a compatible installed Windows SAPI voice, with an online Edge TTS fallback when needed.
- Choose a compatible voice.
- Use Speak to turn reply audio on or off.
- Remember that fallback speech may require internet.
Screen understanding
The optional local vision model can analyze a screenshot of the desktop or a specified window and answer questions about visible text and controls.
- Install the vision model and projector.
- Specify the target window clearly.
- Avoid capturing sensitive information unnecessarily.
Browser awareness
With a supported browser setup, the agent can inspect interactive elements and operate a visible Edge or Chrome browser session.
- Test the browser integration.
- Make the target website explicit.
- Review any submission, purchase or deletion action.
What is on screen can become part of the task.
Ask the vision tool to describe a window, read visible labels or help locate a control. The documented flow captures a screenshot and downscales it for processing, so tiny text and fine details may not always be legible.
- ✓Ask focused questions, e.g. “What error message is visible?”
- ✓Capture only the application or information needed.
- ✓Verify critical text manually before taking action.