Superbash
SearchUser Hub

Hermes Agent · Tutorial 24

Talk to Hermes and Continue on Your Phone

Use voice mode and a local wake listener, then hand the same session to a configured messaging platform without losing its transcript.

Hand-cut paper microphone and local listening dial passing a conversation ribbon from laptop to phone
Reading time
12 min
Last updated
August 2026

0 of 1 complete

Reference guide · Read when needed · View the route

Last tested and updated: August 25, 2026, with Hermes Agent v0.20.5. Run /version first.

Typing is awkward while cooking, walking, or inspecting something with both hands. Hermes can run a spoken session, listen locally for a wake phrase, and hand the same conversation to a configured messaging platform when you leave your desk.

These features have more setup and privacy implications than ordinary slash commands, so start with a harmless test conversation.

/voice: a spoken session

/voice

Voice mode uses speech-to-text and text-to-speech providers configured in Hermes. It supports streaming and barge-in, but the actual microphone, transcription, model, and voice pipeline depend on installed extras and provider configuration. You may need PortAudio, FFmpeg, and voice/STT/TTS credentials.

Test with something non-sensitive:

Help me make a hands-free grocery checklist. Ask one question at a time and read back the final list.

Check that interruption works, names are transcribed correctly, and the final text transcript matches what you intended before using voice for consequential work.

/wake: listen before the conversation starts

/wake on
/wake status
/wake off

Wake listening is off by default and supported on CLI, TUI, and Desktop. The wake detector can run locally, but that does not mean the complete voice pipeline is local: transcription, model inference, or speech generation may still use configured cloud providers.

Only one process should own the microphone. If wake or voice cannot start, check for another Hermes surface or application using the input device.

/handoff: move the same session

From the CLI:

/handoff telegram

Handoff sends the full session—including transcript and tool-call history—to a new thread on a configured platform when possible. It requires the gateway and a home destination configured with /sethome. Hermes refuses to hand off in the middle of a turn, and the CLI exits after a successful transfer.

A beginner test flow

  1. Create a disposable session and ask for a weekend packing list.
  2. Turn /wake on, activate it, and add one item by voice.
  3. Read the transcript and correct any recognition error.
  4. Configure a private messaging home and gateway.
  5. Use /handoff telegram only when Hermes is idle.
  6. Continue the same list from the new thread and confirm earlier context is present.

What users say

Miss Sentient highlighted 24/7 phone access as a major advantage in a one-week test, while also saying browsing was inconsistent and declining to connect Notion or Gmail for security reasons. That is a balanced beginner lesson: mobility is useful without turning every account into an agent tool.

Public X post from Miss Sentient describing one week of Hermes use with mobile access benefits and security concerns
Authentic public post captured August 25, 2026. This is one person’s experience, included to show both usefulness and caution.

Sudo su praised phone and messaging access alongside scheduling, model choice, delegation, memory, and skills. These are individual experiences, not evidence that your network, provider, wake detection, or transcript delivery will be equally reliable.

The stable CLI guide and slash-command reference are the current behavior baseline.

How to know it worked

  • Voice produces an accurate visible transcript and supports interruption.
  • Wake status shows the listener state and releases the microphone when turned off.
  • Handoff occurs only while idle, exits the CLI, and opens the intended private thread.
  • The destination contains the correct prior session—not another recent chat.

Frequently asked questions

Is voice mode entirely local?

Not necessarily. The wake listener may be local while STT, the model, or TTS uses cloud providers.

Why does /voice fail to start?

Check optional dependencies, audio permissions, FFmpeg/PortAudio, provider credentials, and microphone ownership.

Can messaging users call /handoff?

The command is CLI-only because its purpose is to leave the CLI for another platform.

Does handoff send only a summary?

No. It transfers the full transcript and tool-call history.

Can I hand off during a running task?

No. Wait until Hermes is idle.

What does /sethome do?

It configures the gateway destination that handoff can target.

Should I use voice for passwords or account numbers?

No. Spoken audio, transcripts, providers, and nearby listeners create extra exposure.