Claude Code + GLM — Usage & Cost

Live token usage, cost & rate-limit quota for Claude Code and GLM (z.ai) — always visible in your VS Code status bar.

🌐 English | 日本語 | 中文 | Русский

Overview

A Visual Studio Code extension that monitors your coding-agent usage in real time — for both Claude Code and GLM (z.ai) — without leaving your editor.

It reads session data from ~/.claude/projects/ locally (no extra network calls) and shows live utilization, cost, and rate-limit quota right in the status bar, with a rich dashboard one click away. Every token is priced per model — Claude and GLM rates are built in — so cost is accurate even when you mix providers in the same session.

[!NOTE] API calls are minimal. The Claude.ai rate-limit call fires only when a JSONL file was recently updated — when you stop using Claude Code, the extension stops calling the API entirely. Each call uses ≈ 9 tokens of claude-haiku-4-5 (≈ $0.00013). Typical cost with default settings: < $0.01 / month. Set claudeStatus.rateLimitApi.enabled: false to stop all new API calls. GLM (z.ai) quota is fetched from z.ai's own monitor endpoint using your existing ANTHROPIC_AUTH_TOKEN.

[!WARNING] Cost figures are estimates. Default rates track publicly announced Claude and GLM pricing at implementation time and may drift. Update claudeStatus.pricing.* / claudeStatus.pricing.models to match current rates (Anthropic pricing · z.ai pricing).

🔀 Two providers, one meter — Claude Code + GLM

This extension treats Claude Code and GLM (z.ai) as first-class providers and auto-detects which one you're using from ~/.claude/settings.json.

Mixed usage is priced correctly — each JSONL entry is priced by its own message.model, so a window that used both Claude and GLM shows an accurate blended cost.

Point Claude Code at GLM (z.ai)

Add to ~/.claude/settings.json:

{ "env": {
  "ANTHROPIC_BASE_URL": "https://api.z.ai/api/anthropic",
  "ANTHROPIC_AUTH_TOKEN": "<your z.ai key>"
} }

The extension then auto-detects z.ai and:

• shows your real z.ai quota — the 5-hour and weekly utilization % from z.ai's subscription dashboard (fetched from …/api/monitor/usage/quota/limit using your ANTHROPIC_AUTH_TOKEN). Falls back to cost-only if no token is found.
• prices each entry by its GLM model using a built-in GLM price table — so cost is accurate instead of Claude's ~5–7× higher rates. Override or add model rates with claudeStatus.pricing.models.

Other Anthropic-compatible providers (AWS Bedrock, direct API key, custom endpoints) are supported too — cost-only, priced from local JSONL. See Provider Compatibility.

Features

📊 Status Bar — Always Visible

Real-time usage summary pinned to the VS Code status bar.

Hover for a detailed tooltip with full token breakdown and reset times.

🗂 Dashboard Panel

Click the status bar item to open a rich dashboard panel with:

• Current Usage — colour-coded progress bars for 5 h and 7 d windows
• Token Cost — 5 h / today / 7 d / month (est.) costs; expandable token breakdown shows per-type counts (input / output / cache read / cache create) with individual costs and cache hit ratio
• Project Cost — per-workspace breakdown (today / 7 days / 30 days)
• Prediction — burn rate ($/hr), time-to-exhaustion, daily & weekly budget tracking; Rate Limit Timeline chart visualises the projected 5 h utilization from now to reset
• Pricing & Settings — always-visible card showing current token pricing rates, provider, API state, and cache TTL; opens VS Code settings with one click
• Usage History — GitHub-style daily heatmap + hourly pattern bar chart

The panel supports light, dark, and high-contrast VS Code themes natively.

🗂 Project-Level Cost Tracking

Automatically maps the open workspace folder to its Claude Code session directory and shows how much you've spent for that specific project — today, this week, and this month.

Multi-root workspaces are fully supported: each folder gets its own breakdown in the dashboard, and the status bar shows the aggregate.

🤖 5h:78% 7d:84% | my-app:$3.21          ← single workspace
🤖 5h:78% 7d:84% | PJ:$5.43              ← multi-root aggregate

🔮 Usage Prediction & Budget Alerts

Based on the last 30 minutes of activity, the extension predicts how long until the 5 h rate limit is exhausted and warns you before it happens.

• Burn rate — current consumption in $/hr (rolling 30-minute window)
• Time-to-exhaustion — estimated minutes until the 5 h window is full, capped at the next window reset time
• Daily / weekly budget — set optional USD caps; progress bars and alerts fire when the configured threshold (default 80 %) is reached
• VS Code notifications — non-blocking warning as the limit approaches, error dialog when it's nearly reached (with "Open Dashboard" action)

Configure via Settings → Claude Code + GLM Usage or the command palette (Claude+GLM: Set Budget…).

📅 Usage History Heatmap

• Daily heatmap — GitHub Contributions-style grid for the last 30 / 60 / 90 days; intensity reflects daily spend; hover any cell for exact date and cost
• Hourly bar chart — average cost per hour of day (last 30 days)

Number of days is configurable via claudeStatus.heatmap.days (30 / 60 / 90).

Requirements

• VS Code 1.73 or newer
• Claude Code CLI with active sessions — the extension reads ~/.claude/projects/**/*.jsonl for token cost data (this path is shared whether Claude Code talks to Claude.ai, z.ai/GLM, Bedrock, or a custom endpoint)

Authentication is optional depending on your provider:

Provider Compatibility

If auto-detection does not work for your setup, set claudeStatus.claudeProvider explicitly in VS Code Settings.

Installation

VS Code Marketplace

Search "Claude Code GLM Usage" in the Extensions panel, or:

code --install-extension max-riabov.claude-glm-usage

Install from VSIX

1. Download the .vsix from the Releases page.
2. In VS Code: Extensions (Ctrl+Shift+X) → ⋯ → Install from VSIX…

Build from Source

git clone https://github.com/MaxRyabov/claude-glm-usage.git
cd claude-glm-usage
npm install
npm run package       # → claude-glm-usage-*.vsix

Usage

The extension activates automatically on VS Code startup (onStartupFinished).

Configuration

All settings are under the claudeStatus namespace in VS Code Settings.

// Example: settings.json
{
  "claudeStatus.displayMode": "cost",
  "claudeStatus.budget.dailyUsd": 5.00,
  "claudeStatus.statusBar.showProjectCost": true,
  // Force GLM/z.ai if auto-detect doesn't pick it up:
  // "claudeStatus.claudeProvider": "z-ai",
  // Add or override a model's rates:
  "claudeStatus.pricing.models": {
    "glm-4.6": { "inputPerMillion": 0.60, "outputPerMillion": 2.20, "cacheReadPerMillion": 0.11, "cacheCreatePerMillion": 0.60 }
  }
}

Credits

This project is based on long-910/vscode-claude-status (MIT) — huge thanks to the original author. It has been rebranded and repositioned around dual Claude Code + GLM (z.ai) support.

Contributing

Contributions are welcome. See CONTRIBUTING.md for setup instructions, architecture overview, and release procedures.

License

MIT