Agentic coding goes hands free as OpenAI brings GPT-Live's full duplex voice control to Codex and ChatGPT on the desktop

OpenAI has expanded the capabilities of its GPT-Live audio AI model by integrating it directly into developer workflows through the ChatGPT desktop application on macOS and Windows. This integration brings full duplex voice control—allowing simultaneous listening and speaking—to agentic systems like Codex and ChatGPT Work, which are accessible within the ChatGPT desktop app. Initially launched on July 8, 2026, GPT-Live introduced a continuous audio model that eliminates rigid turn-taking by delegating complex reasoning to background models such as GPT-5.5. The latest update extends this conversational interface to technical tasks, enabling software engineers to orchestrate multi-threaded coding jobs, review pull requests, and debug applications using natural voice commands. This advancement could facilitate hands-free software development and collaborative coding sessions for Codex’s over 5 million weekly active users. Codex, OpenAI’s coding-focused model suite, has evolved into a broader productivity platform this year. According to OpenAI, this is the first time voice activation has been natively integrated with Codex on desktop platforms. The integration decouples the real-time voice layer from execution engines, allowing GPT-Live to maintain fluid conversations with verbal acknowledgments while passing heavy computational tasks to background reasoning models. On macOS, the desktop app includes "Appshots" and screen context features that enable ChatGPT Voice to analyze the active window, local files, codebase structures, and plugins. This setup supports a pair-programming dynamic where developers converse naturally while agents execute tasks asynchronously. Developers can issue multiple concurrent voice commands to manage tasks across Codex and ChatGPT Work environments. For example, a developer can simultaneously instruct the system to investigate bugs, review pull requests, and generate unit tests. The application coordinates these actions across Slack conversations, GitHub repositories, and local codebases. It also supports multi-folder projects (build 26.715) and remote execution via iOS, allowing engineers to monitor progress and redirect tasks without switching applications. The voice-enabled desktop release operates under a proprietary commercial model, available to paid subscribers on Plus, Pro, Business, Enterprise, and Education plans. The underlying model weights, voice processing pipelines, and agent state architectures remain closed, with no option for self-hosting or modification. Voice-triggered tasks consume standard usage quotas from existing Codex and ChatGPT Work plans. Developer communities have responded positively to the update, highlighting the potential for hands-free orchestration of complex coding workflows and remote build management. The release has been noted as a step closer to more autonomous and interactive AI-assisted software development.

Key topics in this update include agentic coding goes hands free, agentic coding goes, and hands free.