OpenAI · 2026-09-17 · notable
Codex CLI 0.155.0 — voice conversations arrive in the terminal
Codex CLI 0.155.0 adds experimental /voice conversations with live transcripts and microphone controls, Touch ID checks on MCP requests, and task hiding, archiving and deletion in the agents overview.
Codex CLI can now be talked to: an experimental voice mode with live transcripts sits in the composer.
What is it?
An experimental /voice conversation mode is the headline of Codex CLI 0.155.0. Turned on through /experimental on supported builds, it opens a WebRTC voice channel with live transcripts, a mute shortcut and recording activity indicators, all in a dedicated strip in the composer. The TUI also gained live reasoning summaries in the status row and completion timestamps once a turn finishes.
How does it work?
Voice input runs over a realtime WebRTC connection and the transcript animates in as speech is recognised, with workspace files linked inside it. On the security side, macOS builds can require Touch ID before an MCP request proceeds, backed by Secure Enclave signing, and the agents overview gained hiding, archiving and deletion of tasks plus worktree ownership details and a confirmation step before a clean managed worktree is removed.
Why does it matter?
Talking to a coding agent removes the typing round trip for the parts of a session that are really dictation — describing a bug, steering a long refactor — and the live transcript keeps a written record of what was said. The MCP Touch ID prompt gives a hardware check on the tool calls an agent makes, which matters as MCP servers reach further into local systems. Teams on Amazon Bedrock also get credentials from a configured command now, with caching and expiry-based refresh instead of a static key.
Who is it for?
Codex CLI users on macOS
Try it
npm i -g @openai/codex@0.155.0, then /experimental to enable /voice