Google and Cursor Launch Two-Way Voice Features for AI Coding Agents
Google's Antigravity and Cursor introduced bidirectional voice modes, enabling developers to interact with AI coding agents via real-time speech during task execution.
US tech giant Google and AI coding tool company Cursor announced on November 7, 2026, that they are adding two-way voice capabilities to their respective coding agents. This update aims to address the friction developers face when monitoring long-running AI tasks by allowing them to ask questions, approve commands, or modify instructions through natural language without returning to the keyboard.
Google introduced "Voice mode" in its agentic software development platform, Antigravity, powered by Gemini Live technology. According to Nick Van der Auwermeulen, a developer relations engineer at Google, the traditional rhythm of working with agents involves typing a prompt, waiting for execution, reading results, and typing again. The new voice mode keeps a live call open in the desktop app and command-line interface (CLI), allowing developers to interrupt the agent mid-task to provide new directions, such as saying "wait, try it the other way."
Cursor launched "Voice Chat with Agents," which runs on xAI's Grok Voice. The company stated that developers can describe changes out loud, ask what an agent is working on, and hear updates as the work progresses. Unlike Google's focus on desktop and CLI, Cursor supports voice chat on the web, iOS, and desktop.
Previously, OpenAI integrated two-way voice into its Codex desktop app, while Anthropic's Claude Code currently only supports one-way voice-to-text dictation. The simultaneous rollout by these major platforms signals a shift in AI programming tools from simple text-prompt interactions toward more natural, continuous conversational collaboration, attempting to mimic the immediate communication experience of human pair programming.