Copilot Meter - Usage & Cost Monitor
Monitor your GitHub Copilot usage — sessions, messages, tokens, and API-equivalent cost — directly from your local VS Code logs. No telemetry. No accounts. 100% private.
Features
| Feature |
Description |
| 📊 Full Dashboard |
Daily/hourly charts, model breakdown, tool call stats, session history |
| 🔲 Sidebar Panel |
Quick glance at today's tokens, cost, sessions, and premium credits |
| 💰 Model-Aware Cost |
Cost calculated using official API rates per model (Claude, GPT-4o, Gemini, etc.) |
| 🎯 Premium Request Tracking |
Counts messages to premium models against your monthly limit |
| 🔢 Token Estimation |
Content-aware fallback (code / prose / mixed) when real counts aren't in logs |
| ♻️ Auto-Refresh |
Data updates every 60 seconds automatically |
| 🖥️ Cross-Platform |
Works on Linux, macOS, and Windows |
Getting Started
- Install Copilot Meter from the VS Code Marketplace
- Click the Copilot Meter icon in the Activity Bar
- Use GitHub Copilot as normal — stats update automatically every 60 seconds
Settings
| Setting |
Default |
Description |
copilotUsage.monthlyPremiumLimit |
300 |
Your plan's monthly premium request limit |
How It Works
CopilotMeter reads Copilot chat session files that VS Code writes locally to your machine — nothing is sent anywhere.
VS Code stores Copilot history in two formats:
- JSON state files — older format with full message history and model info
- JSONL request logs — newer format with per-request data including real completion token counts
Both formats are read automatically.
Understanding the Numbers
Sessions
One session = one continuous Copilot chat conversation. A new session starts each time you open a new chat or VS Code creates a new context. Session counts are always exact.
Messages
One message = one prompt you sent to Copilot. Assistant replies are not counted.
A conversation with 3 back-and-forth exchanges → 3 messages (not 6)
Premium Requests
One premium request = one user message sent to a premium model.
Premium models (count against your monthly limit):
- Claude Sonnet, Claude Opus, Claude Haiku
- GPT-4o, GPT-4, o1, o3
- Gemini 1.5/2.0/2.5 Pro, Gemini Flash
- Mistral Large
Base models (free, do not count):
- Default Copilot model (GPT-3.5 / base tier)
5 messages with Claude Sonnet + 10 with the base model → 5 premium requests
Tokens
| Type |
Source |
Accuracy |
| Output tokens |
Real counts from JSONL logs (when available) |
Exact |
| Output tokens (fallback) |
Estimated from response text length |
±5–10% |
| Input tokens |
Always estimated from prompt text length |
±5–10% |
Estimation method (based on cl100k_base encoding used by GPT-4/Copilot):
| Content type |
Chars per token |
| Code |
3.0 |
| Mixed |
3.5 |
| Prose |
4.0 |
Enterprise / Business users: Real output token counts may not appear in local logs. All token figures will be estimated, but sessions, messages, models, and tool calls are always exact.
Estimated Cost
Cost is the API-equivalent value of your usage — what you would pay if billed directly via the model's public API. This is not your actual Copilot subscription cost.
Official rates used (per 1M tokens):
| Model |
Input |
Output |
| Claude Opus 4 |
$15.00 |
$75.00 |
| Claude Sonnet 3.5 / 4 |
$3.00 |
$15.00 |
| Claude Haiku |
$0.80 |
$4.00 |
| GPT-4o |
$2.50 |
$10.00 |
| GPT-4o mini |
$0.15 |
$0.60 |
| o1 |
$15.00 |
$60.00 |
| o3 |
$10.00 |
$40.00 |
| o3-mini / o1-mini |
$1.10 |
$4.40 |
| Gemini 2.5 Pro |
$1.25 |
$10.00 |
| Gemini 2.0 Flash |
$0.10 |
$0.40 |
| Gemini 1.5 Pro |
$1.25 |
$5.00 |
| Mistral Large |
$2.00 |
$6.00 |
Example: 100,000 input + 20,000 output tokens with GPT-4o
= (0.1 × $2.50) + (0.02 × $10.00) = $0.45 estimated cost
Real-World Example
You use Copilot in a coding session:
- Ask Claude Sonnet to explain a function → 1 message, 1 premium request
- Ask it to refactor the code → 1 message, 1 premium request
- Ask a follow-up question → 1 message, 1 premium request
Dashboard shows:
- Sessions:
1
- Messages:
3
- Premium Requests:
3
- Tokens: estimated from message content
- Estimated Cost:
$0.000x (model-aware)
Privacy
- ✅ 100% local — reads only your machine's VS Code storage files
- ✅ No telemetry, no network requests, no accounts required
- ✅ No data ever leaves your machine
Made by OSHIT
