Skip to content
| Marketplace
Sign in
Visual Studio Code>Visualization>Claude Usage CrabNew to Visual Studio Code? Get it now.
Claude Usage Crab

Claude Usage Crab

saegim

|
651 installs
| (2) | Free
See which Claude Code subagents are running, and spot the one that's stuck. Your agents as pixel characters in the sidebar, plus live session, weekly and context usage. Unofficial, read-only.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Claude Usage Crab

Agents tab: a subagent working at the desk in an isometric room, with the Crew roster of pixel characters below

See which Claude Code subagents are running, right now, without leaving VS Code. Every agent in .claude/agents/ becomes a distinct pixel character. When one is actually invoked it walks into the Room, sits at the desk and gets to work. Live session, weekly and context usage sit in the same panel and in the status bar.

Read-only, hook-free, local. Nothing to configure.

Unofficial community extension. Not affiliated with Anthropic.

Agents

A background subagent is a black box. You fire one off and then wonder whether it is working, finished, or wedged. This tab answers that at a glance.

An agent typing at the desk while the full Crew roster shows every agent with its level    A different room theme with the roster collapsed into a compact now-running strip

Agents walk in when invoked and type at the desk while they work. Fold the Crew and it becomes a compact "now running" strip with the task and elapsed time. Pick from twelve rooms in a thumbnail grid.

  • Who is running, as a character in the room, not a log line you have to go looking for. Parallel calls of the same agent stand side by side as separate characters.
  • What it cost. The harness stamps each finished subagent's tokens, tool calls and duration into the transcript, and the panel reads them back. "Which agent eats my weekly limit" is a fact here, not a guess.
  • Stuck? An agent running far past its own usual duration is flagged amber instead of quietly vanishing, so a wedged background agent is something you notice.
  • Levels that mean something. Finishing work earns XP, so an agent's level is a readout of how much it has actually done for you. Your history is swept once on first run, so nobody starts at Lv.1 with a year of work behind them.

Hovering a character shows a crew card with its level, XP progress, lifetime tasks, tokens and active time

Hover a character for its crew card: level and XP progress, plus the lifetime tasks, tokens and active time that earned it.

Per-agent settings with a nickname, role, model dropdown and the human character picker    The same settings sheet showing the monster character picker

Give an agent a nickname and a role, pick its character from 38 sprites, or change its model. The model change is written straight to the agent's .claude/agents/*.md. Agents you never want to see can be hidden, and brought back from the Hidden row.

Detection is transcript-based and read-only. It never blocks or changes how your agents run, and it needs no hooks or setup. Claude Code's own built-in agents (Explore, Plan, general-purpose) show up too, labelled as such, so a character you never created is never a surprise.

Because a background subagent has no knowable progress, running agents show an indeterminate bar and elapsed time, not a fabricated percentage.

Usage panel

Usage panel with session, weekly, Fable and context meters, a usage ring, the crab mascot and today's stat tiles

idle crab    working crab    despairing crab    stunned crab

The crab reads your session usage: calm under 50%, heads-down at 50%, unravelling past 70%, flat-out near the limit.

Weekly details

Weekly details: this PC's share of the account percentage, the other PC on the same account beside it, then tokens and requests, a seven-day bar chart with subagents stacked, and folds for project, model, agent and session

Press Details on the Weekly meter for this PC's last seven days: tokens per day with main sessions and subagents stacked, requests, and an API-equivalent figure, then where it went, by project, by model, by agent and by session.

One Claude account, several PCs. If you share an account, the weekly percentage is the account's, so your own machine's share of it is not knowable from one PC. Turn on Share in that same Details sheet and every PC signed in to the same account that also agreed shows up side by side, adding up to the account's weekly percentage. A row is that percentage divided by what the PC spent, measured at API-equivalent prices, so the table reads the same on every PC on the account. Because the rows divide the whole account total, use nobody there reports (the web, the apps, a PC that did not agree) is spread across them rather than shown apart.

There is no code to type. The group is the account itself, identified only by a hash of its id, so only PCs signed in to it can read the group. It is off until you agree, and agreeing is per account: if you sign in to a second account, that one is off and asked separately, because the people who would see your row are different ones. The question is not asked where VS Code telemetry is off, and claudeUsage.team.share turns the account you are signed in to on or off at any time. Exactly what leaves the PC is listed under How it works and privacy.

Credits and games

The credits balance under the panel title, and the blackjack table below it: a crab dealer, a pixel player, cards on dark green felt and the stake buttons   High-Low: one card face up on the felt, with the price of calling higher and the price of calling lower read off the cards left in the shoe

Slots: three reels stopped in a window cut to one row, with the paying line lit   Wheel: the pointer resting on a slice, every slice carrying the multiplier written on it

Coin Market: a candle chart for one coin, the position's profit and how many coins it holds, and an amount box that takes either credits or coins

The tokens you spend earn credits: one per 10K, weighted the way the API prices them, so output counts five times an input token and a cache read a tenth. On first run your past sessions are counted once in the background, so you start with what you have already earned. The balance sits under the title, next to a coin that fills toward the next credit.

Click it for the games list, where the crab runs five games. Star the ones you come back to and they move to the top.

Card tables. Blackjack: six decks, the dealer stands on soft 17, blackjack pays 3:2, double on your first two cards, split once. High-Low: one card lies face up and you call whether the next one beats it, with each side priced off the cards actually left in the shoe, so the odds move as the run goes on.

Machines. Slots: three rings of 24, rolled to the position the extension picked, so the reel you watch is the reel that paid. Wheel: every slice carries its own multiplier, written on it. The wide ones pay nothing and the thin one pays twelve.

What you keep. Coin Market: four coins (Tidether, Kelpereum, Shelliba and Crabidia) on a candle chart you pick the window of: three hours, a day, a week or a month. Buy with any amount of credits, typed in credits or in the coin itself (half a coin is fine), buy the same coin again as often as 16 times, sell whenever you like, 0.5% on the round trip, and the position is still there when you come back. The fee is small because the holding is long: a coin's realised move is 1.5% to 3.2% in a day, so coming back tomorrow is worth more than closing costs, while a second of it is never worth the trip.

Nothing you do moves a coin's price. The chart is the same in every window and on every install, it is drawn from the instant it belongs to, and there is no way to buy into a rise you have already seen. The price moves on a grid of a 64th of a minute, so the panel has something new to show every second it is open, and the figure it shows is the one the extension will trade at. Every game with a pay table returns between 96% and 97%, and no symbol line, slice or stopping rule beats another; each table was solved by counting every outcome rather than sampled. The market has no table and no round, so there is nothing to return a percentage of: what is proved about it instead is that no rule for deciding when to sell is worth more than any other, and the round trip costs 0.5%.

The extension deals and judges, so a hand survives switching tabs; reloading the window voids it. Your coin positions are what is kept. Credits are earned by using Claude Code and nothing else. They cannot be bought, and they cash out to nothing.

Features

  • Agents. Your subagents as distinct pixel characters that walk into a room when invoked, with levels, lifetime stats and per-agent nickname, role, character and model. Read-only, hook-free.
  • Credits and games. Your token spend becomes credits, and the crab runs five games for them: two card tables, two machines and a coin market. Hide the whole thing from the settings gear if you would rather not have it.
  • Session (5h) / Weekly (7d) / Fable (7d). The same live numbers as Claude Code's built-in usage dialog.
  • Weekly details. Seven days of this PC broken down by project, model, agent and session, and, if you share an account, every PC on it side by side so you can see your own share of the weekly limit. Off until you agree.
  • Context window. How full the current conversation is, scoped to the workspace you are actually in, with the window tier detected per session.
  • Today's tokens and requests. Summed across your local Claude Code sessions, with a +N k delta after each response. Resets at local midnight.
  • Status bar. A compact 5h 54% · wk 46%, turning red as you approach the limit.
  • Burn-rate warning. A heads-up when one recent request's token spend spikes well past today's average.
  • Fold what you don't need. Room and Crew each collapse from their own header and remember the choice, so the panel takes only the space you want to give it.

How it works and privacy

Everything is read locally from Claude Code's own configuration folder. No telemetry, no third party, and no server of ours unless you agree to share with your account's other PCs (below).

  • Which folder. ~/.claude, unless CLAUDE_CONFIG_DIR moves it, which is the usual way to keep a second account. It is resolved in this order: the claudeUsage.configDir setting, CLAUDE_CONFIG_DIR in the Claude Code extension's claudeCode.environmentVariables (where a VS Code profile sets it, and which never reaches the extension host as a real environment variable), CLAUDE_CONFIG_DIR in the environment, then ~/.claude. A path that does not exist is passed over, with the reason in the Claude Usage output.
  • Agents, tokens, requests, context are parsed from your local session transcripts (projects/**/*.jsonl in that folder), incrementally, so a long day of transcripts does not mean re-reading them every tick.
  • Session, weekly and Fable limits come from listening rather than polling. When the Claude Code extension checks its own usage in the same host process, this extension observes that response through Node's diagnostics_channel and reuses it. That means zero extra API requests in normal use, and numbers that match the built-in dialog exactly because they are the built-in dialog's numbers.
  • Fallback. If that tap has been quiet for a while, the extension calls Anthropic's usage endpoint (https://api.anthropic.com/api/oauth/usage) itself, using your own signed-in token, or Claude Code's stored OAuth token if you have not signed in.

Your token is sent only to api.anthropic.com and nowhere else. The code is open, and it is plain JavaScript with no build step, so what ships in the .vsix is what is in the repo.

Sharing with the same account's PCs (off until you agree). For people who share one Claude account. Once per Claude account, the extension asks whether to share; say yes (or choose Share under Weekly > Details, or set claudeUsage.team.share) and every PC signed in to the same Claude account that also agreed shows up side by side, adding up to the account's weekly %. The answer belongs to that account. What you agree to is who sees your row, so signing in to another account (a different configuration directory, above) starts off and asks again there; an answer is never carried from one account to another. The answers are kept under the same hash of the account id that the server sees, so agreeing to share does not put an account id in the extension's own storage either. No code to type: the group is the account, identified only by a hash of its id, so only PCs signed in to it can read it. While sharing is on, every 5 minutes this PC sends the extension author's server (saegimai.up.railway.app): your display name, the weekly window, the account's weekly %, the plan type (Max 20x, Max 5x or Pro), and this PC's use in that window (tokens and an API-price equivalent, per model). The server also keeps one snapshot of these per PC per hour, so the estimate can be checked over a week. Never prompts, file or project names, session titles or any token. A PC that does not agree sends nothing, and the question is not asked where VS Code telemetry is off. Stop sharing under Weekly > Details.

The usage endpoint and the transcript format are both undocumented and may change. If they do, the affected numbers may stop updating until the extension is updated.

Getting started

Install, then open the Explorer sidebar. The Claude Usage panel appears alongside your files with Usage and Agents tabs. The status bar item shows a compact summary and opens the panel on click.

Don't want it on the left? Drag the Claude Usage view header into the secondary side bar on the right, or down into the bottom panel next to the Terminal. VS Code remembers where you put it.

Agents appear automatically for any .claude/agents/*.md in the workspace or in agents/ inside Claude Code's configuration folder (~/.claude/agents/ unless CLAUDE_CONFIG_DIR moves it). No hooks, no config.

Requirements

  • Claude Code (CLI or VS Code extension) signed in with a Claude account.
  • macOS, Linux or Windows. Data is read from ~/.claude, or from wherever CLAUDE_CONFIG_DIR or claudeUsage.configDir points.

License

The extension source is MIT. See LICENSE.

The bundled pixel art is not covered by the MIT licence. It is licensed third-party artwork included here for use within this extension only, and is not offered for reuse or redistribution. See the third-party notice in LICENSE.

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