UsageBar

Windows-first tray app for tracking AI coding subscription usage across providers in one place.

Download

Windows alpha/beta builds are published as GitHub prereleases. Download the latest Windows NSIS installer from UsageBar releases. Release process and preflight checks live in docs/releasing.md.

Install

1. Download and run UsageBar_*_x64-setup.exe.
2. Open UsageBar from the Start menu or tray.
3. In Settings, enable a provider and follow its setup instructions.

Providers

Current Windows rollout status comes from each provider's plugin.json manifest in this fork.

Status meanings:

• Supported: intended to work on Windows from the documented setup path.
• Experimental: visible and testable on Windows, but setup, API shape, or live-account validation can still change.
• Scope: names the usage source. Provider-reported means the provider API or local provider database supplies the number directly. Estimated or telemetry-based means UsageBar derives it from local logs, quota counters, known pools, or manually supplied sessions. Each provider page contains the exact source and limitations.

Provider	Windows status	Scope
Abacus AI	Experimental	API-key usage and credit details
Alibaba	Experimental	Coding Plan request quotas and Bailian Token Plan credits
Amp	Experimental	Free tier, bonus, credits
Antigravity	Supported	All models
Augment	Experimental	Credits via signed-in Augment web Cookie header
Claude	Supported	Session, weekly, extra usage, local token usage (ccusage)
Codebuff	Experimental	Credit balance and weekly rate limit via API token or codebuff login credentials
Codex	Supported	Session, weekly, reviews, credits, managed multi-account selection
Copilot	Experimental	AI Credits, legacy premium requests, chat, completions
Cursor	Supported	Credits, total usage, auto usage, API usage, on-demand, CLI auth
Chutes	Experimental	Rolling 4-hour and monthly subscription quota
DeepSeek	Experimental	API balance with paid and granted credit breakdown
Devin	Experimental	Daily and weekly organization quota
Doubao	Experimental	Volcengine Ark request-limit headers from a minimal probe
Factory / Droid	Experimental	Standard and premium usage buckets
Gemini	Experimental	Gemini quota buckets and reported Code Assist tier
Grok	Experimental	CLI-authenticated billing credits and pay-as-you-go cap
JetBrains AI Assistant	Supported	Quota, remaining
Kilo	Experimental	Direct API-key usage endpoint
Kimi Code (Moonshot)	Experimental	Kimi CLI, kimi.com membership, session and weekly quota from local kimi login OAuth; optional official Moonshot API balance via https://api.moonshot.ai/v1/users/me/balance
Kiro	Experimental	Credits, bonus credits, overages tracking
MiniMax	Experimental	Coding Plan session usage, explicit reported plan when available
Mistral	Experimental	La Plateforme usage and billing details via official Admin API key, with session-cookie fallback
Ollama	Supported	Plan, session, weekly
OpenCode	Supported	OpenCode Go account-wide 5h, weekly, and monthly quota from the official usage API; local history and optional OpenCode Zen balance
OpenAI API	Experimental	Organization API spend windows, completions tokens, requests, and top model from the OpenAI Admin API
OpenRouter	Experimental	Credits, balance, request-rate detail
Perplexity	Experimental	Recurring, purchased, and bonus credit pools via manual cookie/env auth
Qwen Code	Experimental	Local Qwen Code token usage and estimated cost history via ccusage
Qoder	Experimental	Big Model credits from the international or China dashboard
StepFun	Experimental	Step Plan 5-hour and weekly quota
Synthetic	Experimental	Direct API-key quota endpoint
Vertex AI	Experimental	gcloud ADC OAuth plus Cloud Monitoring quota usage
Warp	Experimental	Request limits and plan badge from an undocumented app GraphQL operation
Windsurf	Experimental	Daily quota, weekly quota, extra usage balance
Zed	Experimental	Dashboard token spend via browser-backed cookie replay, with local telemetry fallback
Z.ai	Experimental	Session, weekly, web searches from undocumented subscription/quota endpoints

Want a provider that's not listed? Open an issue.

What It Does

UsageBar sits in your Windows tray and gives you one quick view of your AI coding usage across providers. It refreshes automatically, supports a global shortcut, and keeps provider integrations modular.

You can also read cached usage locally through the HTTP API or CLI, review supported history/reporting, manage credentials, and configure quota notifications.

Current Limitations

UsageBar v0.1.1 is a Windows-first public release.

• Windows installers are unsigned until Authenticode signing is configured. Windows can show Unknown publisher or SmartScreen warnings.
• We welcome provider feedback about setup, usage data, and account support.
• Windows is the primary tested platform for this fork. macOS and Linux remain secondary until the Windows release path is boring.
• Provider coverage is uneven: Supported means the Windows path is intended to work; Experimental means setup, API shape, or live-account validation may still change.
• Some providers report usage directly; others estimate from local history, known quota pools, telemetry logs, or manually supplied session cookies. Provider docs describe the source per integration.
• Signed updater metadata is the primary update path. UsageBar verifies the published asset digest, downloads the Windows installer, then restarts after the app exits.
• Authenticode-signed Windows artifacts, live Edge-account validation, and full crash-recovery expectations remain future work.

Architecture

UsageBar is a Tauri v2 desktop app with a Rust host and a React/TypeScript frontend. Provider integrations live as JavaScript plugins under plugins/ and are copied into the Tauri resource bundle for desktop execution.

• Rust host: tray/window lifecycle, local HTTP API, updater, credential storage, SQLite access, and guarded plugin host APIs.
• React frontend: tray panel, Settings window, provider setup, usage views, preferences, and update prompts.
• Plugin manifests: provider identity, platform support, icons, docs links, and capability declarations.
• Bundled plugins: generated by bun run bundle:plugins before dev/build so desktop resources match source plugins.

Uninstall, Data, Privacy And Security

Uninstall UsageBar in Windows Settings > Apps > Installed apps > UsageBar > Uninstall. If a local test build was installed manually, rerun its installer and choose uninstall if Windows does not list it yet.

UsageBar is local-first. App settings, provider order, display preferences, and app-owned provider secrets live under %APPDATA%\com.sunstory.usagebar; Windows secrets are encrypted with DPAPI. Some providers also read their own local CLI, IDE, browser, or cloud SDK files, documented on each provider page. Legacy beta installs may still use %APPDATA%\com.sunstory.openusage; do not delete it until migration is verified.

• Secrets stay on the machine unless a provider plugin must call that provider's API to read usage.
• UsageBar does not send provider credentials, raw usage payloads, API keys, cookies, or app-owned provider secret files to UsageBar-owned services.
• Plugin host APIs are allowlisted and capability-gated for sensitive operations such as write-capable SQLite access.
• The WebView uses a restrictive starter content security policy.
• The opt-in local HTTP API binds to 127.0.0.1:6736 and requires a bearer token when enabled.
• Telemetry uses the app's analytics integration only for product diagnostics; provider usage payloads and credentials are not telemetry data.
• Crash logs are local support artifacts under %LOCALAPPDATA%\com.sunstory.usagebar unless a user explicitly attaches sanitized logs to a report. Automatic crash upload is not part of the Alpha 1 promise.

Contributing

• Experimental providers. Since they are in an experimental phase, we welcome feedback on your experience.
• Add a provider. Each one is just a plugin. See the Plugin API.
• Read usage locally. See the Local HTTP API.
• Use terminal output. See the UsageBar CLI.
• Fix a bug. Keep the change small, focused, and verified.
• Request a feature or report a bug. Open an issue. Include the provider, auth source, Windows-specific constraints, app version, and sanitized logs. See bug report notes.
• Share diagnostics safely. Include exact error text and timestamps, but do not attach API keys, cookies, raw credential files, or provider-secrets.json.

Keep it simple. No feature creep, no AI-generated commit messages, test your changes.

Lineage

UsageBar started from the OpenUsage codebase. This fork also borrows practical Windows ideas from CodexBar and provider reference patterns from ccusage where they fit.

License

MIT

---

Build from source

Warning: The main branch may not be stable. It is merged directly without staging, so users are advised to use tagged versions for stable builds. Tagged versions are fully tested while main may contain unreleased features.

Stack

• Tauri v2
• Rust
• React 19
• TypeScript
• Vite
• Tailwind CSS v4
• Zustand
• Vitest

Local quality checks

bun run check
bun run test -- --run

bun run check is the default local quality gate for frontend and Rust formatting, ESLint, TypeScript typechecking, and Clippy. The narrower bun run check:frontend and bun run check:rust scripts are available when working on only one side of the application.

Local release build

For a Windows release build on this machine:

bun run release:check -- --release-tag v0.1.1
bun run build:release -- --bundles nsis

If TAURI_SIGNING_PRIVATE_KEY is unset, the helper automatically adds --no-sign so the local build can skip Tauri updater signatures. Set USAGEBAR_ALLOW_UNSIGNED_WINDOWS_INSTALLER=1 for local builds without an Authenticode certificate. The helper signs the final setup executable when Windows signing material exists. The setup executable lands under src-tauri/target/release/bundle/nsis/.

GitHub publishes unsigned Windows installers until the project gets an Authenticode certificate. These installers can show Unknown publisher and trigger Windows SmartScreen's "unrecognized app" warning. Stable releases still require signed Tauri updater metadata; see docs/releasing.md.

Before pushing a release tag, run the same preflight with --require-clean so the tag is cut from a clean worktree.