In the IDE I can add a folder mid-session (File → Add Folder to Workspace), save as .code-workspace, and keep working with the same agent/chat.
In Agents Window, if I need a second repo after planning/implementation started on one repo, editing the workspace creates a new workspace identity. The active agent stays on the original single-repo workspace. I must start a new agent and manually carry context (@Chats / export).
Expected
Add or remove repos from the current Agents Window workspace without creating a new workspace identity
Keep the same agent session and chat history when expanding to multi-root
Parity with IDE behavior
Impact
Common when discovery mid-task reveals a dependency on another repo (frontend + API, app + shared lib). Forcing a new agent breaks flow and loses seamless context.
Workaround today
New agent on multi-root workspace + @Chats or exported plan.
Yep, that’s right. Right now, Edit Workspace on a single-folder workspace creates a new workspace identity, and the active chat stays tied to the original folder. Keeping the same session when expanding like this is something the team is considering, but I can’t share a timeline.
As a workaround, you can try asking the agent to switch directly to a multi-root workspace instead of editing the workspace via the UI. For example: “move this agent to a workspace containing /path/to/frontend and /path/to/api”. The agent can switch to multi-root on its own and carry the conversation over, so you shouldn’t need to start a new chat.
One note: this path isn’t always smooth yet. In some cases the chat can get duplicated or a turn can get interrupted, so it’s a good idea to keep an export of the plan handy. If you run into anything weird, send the Request ID and we’ll take a look.