Claude Code VS Code Adds a Subagent Agent Map

Claude CodeView original changelog

Claude Code's VS Code extension added an agent map in version 2.1.269, giving users a visual overview of every subagent a session has spawned instead of scrolling through a transcript to find them. An "N agents" pill in the footer opens the map, which shows a card for each subagent with a Stop agent action and a read-only transcript, and Focus view now shows live progress rows for running subagents nested under their tool-call groups. The changes make multi-agent sessions in VS Code easier to monitor and intervene in without losing track of what each delegated task is doing.

Key Takeaways

  • An "N agents" footer pill now opens a full map of a session's subagents, replacing transcript-scrolling with per-agent cards.
  • Each subagent card includes a Stop agent action, letting a user cancel one delegated task without disrupting the rest of the session.
  • Read-only transcripts on each card let a user inspect exactly what a subagent did without switching into it.
  • Focus view now shows live progress rows for running subagents, nested under the tool-call groups that spawned them.
  • The feature targets a common pain point of multi-agent orchestration becoming unmanageable once more than a couple of subagents are running at once.
  • It arrives alongside the new VS Code Hooks and Permission rules dialogs in the same release, part of a broader push to make session internals visible and editable without leaving the editor.

From a Wall of Text to a Map

Claude Code's subagents let the main session delegate side tasks to fresh, independently-scoped Claude instances that report back a summary rather than flooding the main conversation with their own work. That delegation is powerful, but until now, keeping track of several subagents running inside one VS Code session meant scrolling through the transcript to find where each one started and what it had done since.

Version 2.1.269 replaces that with an agent map. An "N agents" pill in the extension's footer opens a map view listing every subagent the session has spawned, each as its own card. From a card, a user can jump into that subagent's read-only transcript to see exactly what it did, or press Stop agent to cancel it without touching the rest of the session.

Live Progress Without Leaving Focus View

The same release adds live progress rows for running subagents directly inside Focus view, nested under the tool-call groups that spawned them. Instead of switching to the agent map to check whether a delegated task is still running, a user working in Focus view now sees that progress update in place as it happens.

Together, the two changes give VS Code users the same kind of at-a-glance visibility into multi-agent work that has so far mostly required watching a raw transcript, making it more practical to hand off several tasks at once and only step back in when a subagent needs attention or has gone off track.

Claude Code VS Code Adds a Subagent Agent Map | Yet Another Changelog