Skip to content

GitHub Copilot for JetBrains adds cloud sandbox and remote access

A November 17 changelog entry puts four JetBrains Copilot features into public preview and ends support for IDE version 2025.2.

Draftship Editorial3 min read

GitHub Copilot for JetBrains adds a cloud sandbox, remote access and browser use

GitHub published a changelog entry on November 17, 2026 covering a GitHub Copilot JetBrains update that puts four features into public preview: a cloud sandbox, remote access to chats, browser use and on-device voice dictation. The same post lists chat interface changes, reliability fixes, enterprise policy behaviour and the end of support for one JetBrains IDE version. The full changelog entry says the cloud sandbox and remote access are being rolled out in phases.

What the four previews do

Cloud sandbox runs Copilot sessions in an isolated environment hosted by GitHub instead of on the developer's computer. There, Copilot can make code changes and run commands and tests. GitHub says this frees local resources and allows parallel sessions.

Remote access is opt in. Once enabled, Copilot chats can appear in other Copilot clients, with VS Code and the Copilot app given as examples. Developers can find shared sessions from remotely accessible computers, or reopen a remote session from history and continue it.

Browser use works alongside GitHub Copilot in Chrome. The built-in plugin can open and navigate pages, read their content, and interact with links, buttons and forms to gather information or finish web tasks. In JetBrains it is switched on under GitHub Copilot > Chat > Enable Browser Use.

Voice dictation runs on device on supported native plugin packages with Editor Preview enabled. You opt in and download a model. Transcripts remain editable drafts until you press Send.

Split mode remote development is also supported: chat and editor UI run on the local client while Copilot operations run on the remote backend, with the plugin installed in both places.

Smaller changes in the chat panel

Enter now steers a running agent session, and Ctrl+Enter (Command+Enter on macOS) queues a follow-up. Copilot Fusion is pinned at the top of Copilot agent model management and stays enabled even when the whole model group is deselected. Session history gets a Delete option beside Archive, plus a way to skip future single-session deletion confirmations, and agent prompts and responses show timestamps. Inline chat shows model names and per-response token details while keeping session credit totals, and Ask mode footers show the selected Fusion tier rather than the underlying model.

Reliability work covers chat persistence, startup, recovery, saved history, proxy loading problems and saved Claude Code paths on Unix, plus clearer errors for failed tool calls and historical session diffs on newer JetBrains IDE EAP builds.

Code review now uses the Copilot agent review engine, with results still in the existing inline review interface. Java modernization findings appear in the Problems View on supported IDEs, and automatic Maven vulnerability checks are back: opening or importing a Maven project sends dependency coordinates to GitHub Advisories.

Enterprise policy and the version cutoff

Enterprise-managed settings now disable Claude and Codex unless an administrator policy allows an exception, and saved provider preferences are preserved. Copilot Fusion applies enterprise-managed tier defaults while keeping available tiers selectable.

Support for JetBrains IDE 2025.2 has ended. Developers need 2025.3 or later to keep receiving Copilot updates, and GitHub recommends 2026.2 or later. The chat onboarding tour is now generally available to eligible users, and usage tips are on by default.

Limits and open questions

GitHub gives no completion dates for the phased rollouts, and the entry does not say which plans or licence types get the cloud sandbox, or what sandbox sessions cost in credits. The headline references Ignite releases, but the body covers the JetBrains plugin only, so what else shipped under that banner is unclear. This suggests the previews are likely to reach different users at different times rather than all at once.

What this means for building apps with AI

For people using AI to write app code, the cloud sandbox and remote access matter most: sessions can move between clients and run off your machine, so an agent can run tests while you keep working. That depends on the phased rollouts. Draftship, which builds and hosts apps from a description, covers its own chat and model choices in using the AI chat and choosing a model.

Primary sources

This article was drafted with AI assistance from the primary sources listed below and checked automatically against them. An editor hasn't reviewed it yet.