· ai-devtools
Cursor, Google join OpenAI in giving coding agents voice chat support
For developers, the ability to speak directly to their coding assistants is becoming a reality, with Cursor and Google joining OpenAI in offering agents a voice, while Claude Code remains silent on…
For developers, the ability to speak directly to their coding assistants is becoming a reality, with Cursor and Google joining OpenAI in offering agents a voice, while Claude Code remains silent on this front.
Both Cursor and Google are championing the idea that developers prefer to converse with their AI agents during the coding process, each unveiling a bidirectional voice capability that allows users to issue requests and provide subsequent guidance without needing to switch back to typing. On Friday, Cursor introduced Voice Chat with Agents, a feature powered by Grok Voice, enabling developers to initiate a call with an agent to discuss a modification. Simultaneously, Google rolled out Voice mode within Antigravity, utilizing Gemini Live to bring spoken interactions into its command-line tool and desktop application.
In a blog post detailing this addition, Nick Van der Auwermeulen, a developer relations engineer at Google, draws a comparison between the fluid back-and-forth of human colleagues and the typical interaction pattern with coding agents. He noted, “Some of the best engineering happens out loud.” Furthermore, he elaborated, “You talk a bug through with a teammate, think out loud in front of a whiteboard, or say ‘wait, try it the other way’ halfway through a pairing session. Working with an agent has mostly had a different rhythm: you type a prompt, the agent works, you read what it did, and you type again.” He continued by explaining, “Some of the best engineering happens out loud. The longer a job runs, he adds, the more often developers need to check in.” These check-ins might involve approving a command or noticing a requirement that the initial instructions overlooked. Van der Auwermeulen explained that “Each of those moments pulls you back to the keyboard and away from whatever you were looking at.” Consequently, Antigravity’s voice mode maintains an active call while the coding agent is operating, allowing the developer to receive updates or approve a command without needing to reach for the keyboard.
Cursor, meanwhile, promotes a similar conversational flow, stating that users can “Describe a change out loud, ask what an agent is working on, and hear updates as the work moves forward.”
To provide context, Antigravity is Google’s platform for agentic software development, which debuted alongside Gemini 3 last November, subsequently receiving version 2.0 in May, which included a dedicated desktop application. This ecosystem also features an Antigravity CLI and extensions for editors such as VS Code, JetBrains IDEs, and Xcode. Google is currently centralizing its various coding utilities within Antigravity, having migrated individual Gemini CLI users to it in June, even as it begins phasing out Jules, a previously existing background coding agent. It is important to mention that Antigravity already possessed a dictation function that converts speech into written prompts. To initiate a two-way call in the desktop application, the user taps the waveform icon adjacent to the dictation feature and then instructs the agent on the desired task. Antigravity will respond verbally and sustain the call throughout the agent's work, permitting the user to inquire about progress or issue further directions. The system also supports interrupting the agent mid-sentence to pose a question or issue another directive, with Antigravity responding audibly. When the agent concludes its task, its written summary and a button to examine the code modifications appear within the same chat interface.
Additionally, this voice mode is functional within the Antigravity CLI; users can commence a call by executing the `/voice live` command, thereby hearing the agent's replies while the work proceeds in the terminal.
Cursor, which was acquired by Elon Musk’s SpaceX for $60 billion in August, has followed a comparable trajectory toward incorporating voice capabilities. After adding dictation in October 2025, the company noted in last month's release notes that existing voice interactions were updated to allow agents to provide brief spoken progress reports. A distinction in Cursor's implementation is the range of access points; Cursor supports voice chat across its web, iOS, and desktop platforms, whereas Google provides it specifically within the Antigravity desktop app and CLI. Cursor’s voice calls operate using Grok Voice, which xAI made available via its API in April.
The prevailing trend, it appears, is the desire to maintain an ongoing dialogue after the agent begins its work. However, Google and Cursor are not the first companies to introduce two-way voice to coding agents. OpenAI released a Codex desktop application in February that included a built-in dictation feature. Furthermore, as reported by The New Stack in July, the company incorporated two-way voice into the ChatGPT desktop application, empowering developers to provide instructions and check on tasks verbally. The Codex CLI has offered live voice conversations since the spring, with OpenAI enabling these by default in September.
Anthropic, for its part, offers a voice mode that provides two-way voice functionality, but this is restricted exclusively to the Claude app. Within Claude Code specifically, the voice interaction remains unidirectional via dictation—the `/voice` command transforms spoken instructions into prompt text, and the agent responds in writing. While dictation itself is useful for verbally describing a bug or outlining a feature, this capability is spreading across the AI coding sphere, reaching from major AI laboratories to well-funded startups like Replit. Ultimately, the current momentum suggests a drive to sustain the conversation after an agent starts working, enabling developers to follow up on decisions, clarify specifications, or alter the course before the task is complete.
Why it matters
This trend signals a major shift in how developers interact with AI coding assistants, moving beyond simple prompt-response cycles. The ability to have live, spoken conversations with agents could significantly change the workflow for debugging and iterative development. Developers should watch how this conversational interface evolves across different tooling ecosystems.
