An AI voice agent for your desktop, not your phone line
Hey Speak is SpeakApp's voice agent: you say “Hey Speak”, ask your question and it answers out loud, speech to speech, with no window to open. It answers itself or passes your question to ChatGPT, Gemini or Claude on the web, then reads you the answer. Nothing to set up: no API key, no account to create. It sits next to the reading loop: your agents' replies are read to you one after another, you say “ok”, answer, and “enter” sends your answer into the right session.
What is a desktop voice agent?
It is the voice assistant of your work machine, the equivalent of the one you already have on your phone. You say “Hey Speak”, ask your question, and it answers out loud. You speak, it speaks: speech to speech, with no interface to look at.
A phone agent answers callers. This one answers you, while your Claude Code, Codex or ChatGPT sessions keep running in their windows.
What can Hey Speak do while you work?
- Answer a question. It replies in the language you asked in.
- Pass your question to another AI. You choose who answers: Hey Speak itself, or ChatGPT, Gemini or Claude on the web. You can also name the platform in your sentence: “Hey Speak, ask ChatGPT...”. It sends the question and reads you the answer when it arrives.
- Transform a text. Select a passage and say what you want: “make it shorter”, “translate into French”, “summarise in three points”. The result is pasted at your cursor, and out loud you only hear “done”.
It does not bluff. When your question needs current news or a search it cannot do reliably, it opens a new conversation on the platform you chose, asks your question once, tells you so, then reads you the answer.
How does it fit the reading loop?
SpeakApp starts with listening. Your agents' replies are read to you one after another, as soon as they are ready. The reading toast, at the top right of your screen, shows one card per waiting conversation, and the widget's circle counts what is left.
To answer the conversation in the toast:
- say “ok”,
- dictate your answer,
- say “enter”: your answer is pasted and sent into the right session.
“Ok” and “enter” are words you say, not keys. You do not hunt for the window. It works in Claude Desktop (the Code tab and local Cowork), the Codex app and ChatGPT in Chrome.
The two work together: the reading loop brings you what your sessions just produced, and Hey Speak answers what you ask it.
What about Copilot?
Copilot is coming. It is the next step of the agent mode: it will follow your agents' work and answer out loud your questions about the sessions in progress, without you reading each window.
What should you know?
The microphone is closed at installation. It opens the first time you use a voice feature, with a button or a shortcut. After that, your voice is enough. If you call Hey Speak and say nothing, it closes after forty-five seconds. It has no long memory: a conversation lasts for the session. And SpeakApp does not read sessions that run in the cloud.
Related guides
Frequently asked questions
Is Hey Speak a phone voice agent?
No. Hey Speak lives on your computer. It answers you, out loud, while you work with your AI agents. It does not pick up any calls.
Do I need an API key or an account to use Hey Speak?
No. You sign in to SpeakApp and it works. No key, no third-party account to create.
Who answers my questions?
You choose: Hey Speak itself, or ChatGPT, Gemini or Claude on the web. Without a choice, ChatGPT answers. You can also name the platform in your sentence.
Does Hey Speak remember our conversations?
No. A conversation of several turns lasts for the session. It has no long memory.
Can I start for free?
Yes, a free plan exists, with every feature and volume limits. The details are on the pricing page.
Try SpeakApp
SpeakApp is available on Windows, coming soon to Mac. Download SpeakApp or join the Mac waiting list. To see the whole product first: What is SpeakApp?