VS Code 1.131 Adds Built-In Dictation and Deeper Subagent Insights

VS Code 1.131 adds two developer-facing features tied to LLM workflows. In the Agents panel, the main chat conversation now displays live details about running subagents — including the model in use, execution time, and actively invoked tools — without needing to open each subagent's own conversation view; users can still switch into a subagent's dedicated chat to see full progress while the main conversation stays available.

The release also bundles an experimental, built-in transcription service for dictation in chat, editor and the integrated terminal, replacing the need for the separate VS Code Speech extension. Microsoft says it uses the private, offline Nemotron model, keeping audio processing on-device. The feature works on Windows x64/Arm64, macOS on Apple Silicon, Linux x64/Arm64 with glibc 2.34 or newer, and in remote workspaces since transcription runs on the local client; it is not yet available in the web version of VS Code, on Intel-based Macs, or on 32-bit/Arm32 systems.

As a lighter addition, typing /vscode-pet in chat now spawns a small draggable virtual pet, echoing the many pet-themed extensions already found in the Visual Studio Marketplace. Alongside 1.131, Microsoft also published the newer Insiders build, VS Code 1.132.