Claude and ChatGPT Voice Modes Now Support Tool Use

Voice Interface Capabilities

OpenAI and Anthropic have both updated their respective voice modes to support tool execution, moving beyond simple dictation. OpenAI has integrated its voice agent into the ChatGPT desktop application, enabling it to function as an orchestrator that can control the computer, such as launching applications or managing system tasks. This implementation utilizes a low-latency, real-time voice model that supports interruptions during processing.

Anthropic has updated Claude to support Model Context Protocol (MCP) servers and external connectors within its voice mode. Unlike OpenAI's desktop-focused approach, Claude's voice functionality is accessible on mobile devices and allows users to select from different models, including Opus, Sonnet, or Haiku. While Claude's voice processing currently relies on a standard speech-to-text and text-to-speech pipeline, which is slower than OpenAI's real-time model, its ability to interface with custom MCP servers allows it to perform complex tasks like calendar management and web research while on the go.

Practical Application

The integration of connectors allows these voice agents to act as functional assistants rather than just conversational partners. Users can now perform tasks such as querying email inboxes, checking calendar bookings via Cal.com, or conducting research using tools like Bright Data, all through voice commands. This shift effectively replaces the need for custom-built, local voice agents that previously required manual orchestration of speech-to-text pipelines and local tool-calling logic.