What is the difference between Compose and Talk?

Compose turns what you say (or a screenshot/file you attach) into a structured, ready-to-run prompt you review before sending. Talk is a live, hands-free voice conversation for supervising your agent — ask it to review, approve, or merge a pull request, catch you up on what it just did or is working on, dispatch a new task, or spin up / wake / stop a Codespace, and it narrates progress back to you.

What can I ask Talk to do?

Talk drives four hands-free supervision tasks by voice: (1) review, approve, or merge a pull request; (2) get a catch-up or status — "what did my agent do?", "what is it working on?", "am I done?" — and Talk reads the session and tells you; (3) dispatch a new task to your agent; and (4) control a Codespace — spin one up, wake it, or stop it. Your agent does the work and narrates progress back to you.

Is Voice Command free?

Compose is free, with a daily cap. Talk is live and is a PRO feature, metered by the minute.

What language does it use?

Compose mirrors the language you speak or type in — no forced translation to English.

Free

Unified Voice Command

One composer button, Compose or Talk — speak to your coding agent from your phone.

01 · COMPOSERReady

Tap Voice

Morphs from Send when the input is empty

02 · CHOOSE MODEYour call

Compose | Talk

One switch, two ways to work by voice

03 · COMPOSELive

Structured prompt

Speak or attach — a Task/Context/Steps prompt appears, editable

04 · TALKLive · PRO

Hands-free supervision

Review/merge a PR, get a catch-up, dispatch a task, or control a Codespace — by voice

Unified Voice Command puts one button in the composer: Voice when the input is empty, Send once you start typing. Tap it and pick Compose — speak or attach a screenshot or file and CodeAgent turns it into a clear, structured Task/Context/Steps prompt in your own language — or switch to Talk to supervise your coding agent by voice, hands-free: ask it to review, approve, or merge a pull request, catch you up on what your agent just did (or is working on), send it a new task, or spin up / wake / stop a Codespace — it works and narrates progress back to you, no typing.

One button, no hunting

The old mic and Smart Compose buttons merge into a single morphing button — Voice on an empty input, Send once you start typing.

Compose structures it for you

Speak or attach a screenshot or file — CodeAgent turns it into a Task/Context/Steps prompt, in your language, before it reaches the agent.

Talk supervises hands-free

Ask your agent by voice to review, approve, or merge a pull request, catch you up on what it just did, dispatch a new task, or spin up / wake / stop a Codespace — it works and narrates progress back.

Free to start

Compose is free with a daily cap; Talk is live for PRO, metered by the minute.

1
Tap the Voice button
Open any session composer — an empty input shows the Voice button.
2
Pick Compose or Talk
A sheet opens with the Compose | Talk switch; your last-used mode is remembered per session.
3
Compose: speak or attach
Record a voice memo or add a screenshot/file — CodeAgent structures it into a ready-to-run prompt you review before sending.
4
Talk: just start talking
Ask it to review, approve, or merge a pull request, catch you up on what your agent did or is working on, kick off a new task, or spin up / wake / stop a Codespace — it works and narrates progress back, hands-free, until you end the conversation.

PRO

Talk is live (app 3.5.0) and is a PRO feature, metered by the minute; Compose works today on any plan (free, with a daily cap). Compose is the evolution of Voice prompts + Smart Composer fused into one entry point — those two pages cover the underlying pieces in more depth.

FAQ

Smart Composer Voice prompts

Ready to try it?

Start free See pricing

Unified Voice Command

FAQ

What is the difference between Compose and Talk?

What can I ask Talk to do?

Is Voice Command free?

What language does it use?

Related