GitHub Copilot update adds side chats, worktrees and local multilingual dictation

GitHub's latest Copilot weekly release is a practical workflow update rather than a single headline feature: it extends the coding assistant across the Copilot app, Copilot CLI and VS Code with more ways to keep parallel work organized. The changelog, published on August 7, says the August 3 release batch is aimed at helping developers resume work, review changes and ask follow-up questions without losing context.

In the Copilot app, GitHub says Auto mode now reports which model handled a completed request and can show AI credit and cache details when those details are available. That matters for teams trying to understand the cost and behavior of agentic coding sessions, especially as organizations put more scrutiny on AI usage. The app also adds direct access to shared sessions and a /side command for opening a parallel question while leaving the main task intact.

CLI and VS Code changes

The CLI side of the release is focused on juggling multiple agent sessions. GitHub describes a new Sessions sidebar with shortcuts to open, close and switch sessions, plus an experimental /worktree command that creates an isolated worktree and starts a separate conversation there. The release also adds /rewind support without Git, intended to restore the conversation and files Copilot changed while preserving later edits, and timeline durations for live tool calls so developers can spot slow-running commands.

VS Code 1.132 picks up several companion features. The integrated browser can now collect element-level feedback, letting a user select and annotate specific page elements before sending visual feedback to an agent. Dictation now uses a multilingual on-device model by default, keeping audio local while following the configured language, browser or system locale, or automatic detection. GitHub also notes a /btw side-chat flow and an experimental hybrid Markdown editor for reviewing Markdown diffs in context.

The update shows how coding assistants are shifting from autocomplete toward managed work sessions. The most significant theme is not just that Copilot can answer more prompts, but that GitHub is adding controls around context, parallel exploration, review, voice input and usage visibility. For engineering teams, those pieces may matter as much as model quality because they determine whether AI work can fit into repeatable development habits.

Sources

• github.blog

Cover photo by Daniil Komov on Pexels, used under the Pexels License.