ModelMeter is a VS Code extension for monitoring AI provider balances and estimating local AI model usage costs across DeepSeek, SiliconFlow, Moonshot/Kimi, OpenAI, Anthropic/Claude, and Zhipu GLM.
Version 1.1.0 is a product-hardening release: it moves the extension to a modular architecture, removes external webview scripts, applies stricter webview rendering, enables custom pricing, and adds manual usage, CSV export, and history clearing workflows.
- Multi-provider accounts: add, switch, edit, delete, and test API accounts.
- Balance monitoring: supported for providers that expose balance APIs.
- Local usage estimation: manually record token usage and estimate cost using built-in or custom pricing.
- Charts without CDN dependency: local chart renderer inside the extension package.
- CSV export: export the active account's local 30-day usage history.
- Privacy-first storage: API keys use VS Code SecretStorage; local history stays in VS Code globalState.
- Proxy support: respects VS Code
http.proxysettings. - Runtime settings: status bar visibility, refresh interval, and custom pricing are applied at runtime.
From a packaged .vsix:
code --install-extension modelmeter-1.1.0.vsixFrom source:
npm install
npm run compileLaunch the extension with VS Code's Extension Development Host.
ModelMeter: Add / Configure API AccountModelMeter: Manage API AccountsModelMeter: Refresh Usage DataModelMeter: Test API ConnectionModelMeter: Add Manual Usage RecordModelMeter: Export Usage CSVModelMeter: Clear Usage History
{
"modelmeter.refreshInterval": 60,
"modelmeter.showStatusBar": true,
"modelmeter.customPricing": {
"openai:gpt-4o": {
"input": 2.5,
"output": 10,
"cachedInput": 1.25
}
}
}Pricing values are USD per 1M tokens.
- API keys are stored through VS Code SecretStorage.
- ModelMeter does not collect telemetry.
- Local usage history is not uploaded.
- Webview scripts are local to the extension package.
- Provider requests are only sent when testing a connection or querying balance.
See SECURITY.md and PRIVACY.md.
MIT. See LICENSE.