Skip to content
| Marketplace
Sign in
Visual Studio Code>Other>Extras for Claude CodeNew to Visual Studio Code? Get it now.
Extras for Claude Code

Extras for Claude Code

Xnerv Wang

|
2 installs
| (0) | Free
Work plan: a tree of what the conversation still has to do, kept current by Claude itself; also adds timestamps, reply figures, navigation, context details, and chimes to the Claude Code chat panel. Patches the installed Claude Code extension's files. Not affiliated with Anthropic.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Extras for Claude Code

This VS Code extension edits the installed Claude Code extension to add a Claude-maintained work plan to its chat panel. The plan keeps unfinished work on screen and puts it back in front of Claude after older turns are compacted. It also adds timestamps; reply duration, context use, running spend, and model; a filterable table of contents with message stepping; a context meter with a hover breakdown; sub-agent tags and filters; a fold for thinking and tool calls; chimes; and session details. Not affiliated with Anthropic. Requires Anthropic's Claude Code extension, which it adds nothing without, and VS Code 1.94 or later. 23 settings, 11 commands.

Install

Search for Extras for Claude Code in the Extensions view, or run code --install-extension xnervwang.claude-code-extras. Anthropic's Claude Code extension installs as a dependency.

Reload the VS Code window twice after installing. Run Extras for Claude Code: Show Status for the patch status. In a Remote SSH window, install on the remote.

Extras for Claude Code edits files inside the installed Claude Code extension. The patch is reverted by removing it or uninstalling the extension; a Claude Code build it does not recognise is left alone with a warning.

Verified against Claude Code 2.1.287; compatibility with other builds unknown.

Work plan

The work plan is a tree of what the conversation you are looking at still has to do, in its own activity-bar view. Compaction summarizes older turns to fit the context window, so Claude can lose track of the main task after several digressions. The plan stays in a file outside the conversation, named for its session ID, and a hook injects the open rows every turn: a task raised twenty turns and several compactions ago remains in front of Claude and visible in the plan's activity-bar view.

Claude keeps the plan current; you do not. The extension ships a Claude Code plugin, installed on first activation, that writes it. Every turn, the plugin puts open rows in front of the model, says something at the end of a turn that changed things without touching the plan, and refuses a row whose description is longer than the view shows. Claude is told when to move a row: a new request becomes discussing; approval makes it todo; starting work makes it doing, before the work; finishing makes it done; a decision against it makes it dropped with the reason; leaving it for later makes it parked with the reason.

If A needs B finished first, B becomes a child of A: a digression sits under the thing it interrupted. Rows have numbers such as 1, 2, and 2.1, so you can name one in conversation without quoting its title. The number is its place in the file; finishing a row leaves a gap rather than renumbering what follows. Unfinished rows appear first, newest first within that. The activity-bar badge counts open rows and is absent, rather than zero, when nothing is waiting.

Each conversation has one file, named by session ID, under ~/.claude/plugins/data/agent-work-plan-claude-code-extras/. The view reads those files; the extension never writes them. Correct a plan by hand from the title-bar button or a row's context menu. A plan belongs to one conversation: in-process sub-agents are not covered, and a detached agent is handed one task and left alone.

What it looks like

The work plan on the left, times on every message and reply, the table of contents at the right edge, and the footer controls. Thinking and tool calls shown, which is the default:

Chat panel with thinking and tool calls shown

The same conversation with one footer button pressed, leaving your messages and Claude's replies:

The same conversation with thinking and tool calls folded away

Features

What it shows Where it appears
Timestamps: transcript times; start→result for tool calls; clock today; day and month for older entries; original times in reopened history Messages, reply blocks, tool calls
Reply figures: 2m10s · ctx 33% · cost $0.42 · opus-5 high; turn duration, context-window use, running spend, model Each reply, including closed turns
Sub-agent tags and filters: spawn descriptions, such as #look up tomorrow's weather; All, Main, and individual sub-agent filters; table of contents follows filter Tags on rows; filters in footer
Conversation navigation: message times, opening words, compaction points; filter box; click-to-scroll; message stepping with arrows, ,, and .; top and bottom jumps Right-edge table of contents; buttons
Context meter: visible even with over half the window free; outline at 15% remaining; category, memory-file, and custom-agent breakdown Chat panel meter; hover tooltip
Session dates: 11d · 09/05→09/16; Claude Code relative last-active time, followed by date span Sessions
Chimes: two rising notes, three knocks, two falling notes; synthesized sound; no shipped audio files; footer mute button Finished turn, permission request, question; local panel, including Remote SSH; footer
Work plan: see Work plan Activity-bar view
Session information: session ID, starting directory, transcript location and size; click-to-copy values Footer button and session information

Settings

Open Extras for Claude Code: Settings for the settings UI. Options cover message times, reply figures, sub-agent tags, navigation, footer controls, sounds, session details, the work plan and its offer thresholds, and panel-opening latency recording. Message appearance: claudeCodeExtras.userMessageColor and claudeCodeExtras.userMessageEdge.

Most settings apply live without a reload; installing or upgrading the patch requires a reload. claudeCodeExtras.enabled is the same setting as the Extras: On/Off status bar item. Turning it off hides the additions without restoring the original files.

Turning off claudeCodeExtras.workPlan hides the view and stops plan updates. The registered plugin's skill description still loads in Claude Code; use Extras for Claude Code: Stop Loading the Work Plan Plugin to stop loading it.

Commands

Command What it does
Extras for Claude Code: Settings Open this extension's VS Code settings
Extras for Claude Code: Toggle On/Off Toggle additions
Extras for Claude Code: Turn On Show additions
Extras for Claude Code: Turn Off Hide additions
Extras for Claude Code: Remove from Claude Code (restore original files) Restore original files
Extras for Claude Code: Show Status Report patch status and what was patched
Extras for Claude Code: Refresh Work Plan Refresh work plan
Extras for Claude Code: Open the Work Plan File Open current plan file
Extras for Claude Code: Install the Work Plan Plugin for Claude Code Register companion plugin
Extras for Claude Code: Show How Long Opening a Panel Took Read recorded panel-opening times
Extras for Claude Code: Stop Loading the Work Plan Plugin Remove plugin from Claude Code settings

Known limits

  • Patched code shapes with renamed CSS classes: panel features may fail; orange exclamation mark at the panel's top right.
  • Uninstallation unregisters the plugin: Claude Code deletes its data directory; plans copied beforehand to ~/.claude/agent-work-plan-plans-<timestamp>/.

Contributing

See CONTRIBUTING.md for the layout, checks, and verified-build records.

License

BSD 3-Clause.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
  • Your Privacy Choices
  • Consumer Health Privacy
© 2026 Microsoft