MacOS menubar app which tracks AI usage through estimated carbon emissions and water & electricity usage.
Swift
2
5 commits
updated Sep 22, 2026
See what people are saying
README
AI Usage Menu Bar
macOS menu bar app. Estimates energy, water, and carbon footprint of your coding-agent CLI usage (Claude Code, Codex, Gemini CLI) using local data only.
Inspired by this MIT Tech Review piece.
How it works
- ccusage reads each tool's local logs and reports daily token usage.
- EcoLogits methodology turns tokens into Wh / mL / gCO2e, based on model size, real latency stats, datacenter PUE/WUE, and grid carbon intensity.
- Coefficients are precomputed at build time (
scripts/build-coefficients.mjs→Resources/coefficients.json), not calculated live. - Only output tokens are counted (decode dominates inference energy).
- Kimi and Grok aren't supported by EcoLogits, so their usage shows in the app but contributes $0 to environmental totals.
Caveat: order-of-magnitude estimate, not a measurement. Unknown/new models get mapped to the closest known model in the same family.
Install
- Download
AIUsageMenuBar.zipfrom the latest release. - Unzip it and drag
AIUsageMenuBar.appto/Applications. - Open it. It's signed with a Developer ID and notarized, so it should launch normally with no Gatekeeper warning.
- It has no Dock icon or window — look for it in the menu bar. From then on it checks for updates automatically via Sparkle.
Build & run
brew install xcodegen # one-time
node scripts/build-coefficients.mjs # regenerate coefficients.json
xcodegen generate # regenerate .xcodeproj from project.yml
open AIUsageMenuBar.xcodeproj # ⌘R to build & run
Generated Xcode 26+ project via XcodeGen — project.yml is the source of truth so re-run xcodegen generate after editing.
Runs ccusage's compiled binary directly (vendored in Vendor/ccusage/). Node is only needed for dev scripts, not for the built app.
Releasing
git tag vX.X.X && git push --tags # triggers .github/workflows/release.yml
That workflow builds, signs, notarizes, publishes a GitHub Release, and updates the Sparkle appcast (docs/appcast.xml) for auto-updates. Auto-updates only work once the repo is public (Sparkle needs public release URLs).
One-time setup for releases (Apple Developer account required) — set these GitHub secrets:
| Secret | Source |
|---|---|
DEVELOPER_ID_P12_BASE64 | Developer ID cert, exported as .p12, base64-encoded |
DEVELOPER_ID_P12_PASSWORD | That export's password |
TEAM_ID | e.g., Z4ZXD95E9B |
NOTARY_APPLE_ID | Apple ID email for notarization |
NOTARY_PASSWORD | App-specific password (appleid.apple.com) |
SPARKLE_PRIVATE_KEY | From Sparkle's generate_keys -x (public half already in project.yml) |
Local sanity check (no notarizing/zipping):
xcodegen generate
xcodebuild -project AIUsageMenuBar.xcodeproj -scheme AIUsageMenuBar -configuration Release build
Similar repos
Languages
Swift
76.1%
JavaScript
21.3%
Shell
2.6%
MacOS menubar app which tracks AI usage through estimated carbon emissions and water & electricity usage.
Swift
2
5 commits
updated Sep 22, 2026
See what people are saying
README
AI Usage Menu Bar
macOS menu bar app. Estimates energy, water, and carbon footprint of your coding-agent CLI usage (Claude Code, Codex, Gemini CLI) using local data only.
Inspired by this MIT Tech Review piece.
How it works
- ccusage reads each tool's local logs and reports daily token usage.
- EcoLogits methodology turns tokens into Wh / mL / gCO2e, based on model size, real latency stats, datacenter PUE/WUE, and grid carbon intensity.
- Coefficients are precomputed at build time (
scripts/build-coefficients.mjs→Resources/coefficients.json), not calculated live. - Only output tokens are counted (decode dominates inference energy).
- Kimi and Grok aren't supported by EcoLogits, so their usage shows in the app but contributes $0 to environmental totals.
Caveat: order-of-magnitude estimate, not a measurement. Unknown/new models get mapped to the closest known model in the same family.
Install
- Download
AIUsageMenuBar.zipfrom the latest release. - Unzip it and drag
AIUsageMenuBar.appto/Applications. - Open it. It's signed with a Developer ID and notarized, so it should launch normally with no Gatekeeper warning.
- It has no Dock icon or window — look for it in the menu bar. From then on it checks for updates automatically via Sparkle.
Build & run
brew install xcodegen # one-time
node scripts/build-coefficients.mjs # regenerate coefficients.json
xcodegen generate # regenerate .xcodeproj from project.yml
open AIUsageMenuBar.xcodeproj # ⌘R to build & run
Generated Xcode 26+ project via XcodeGen — project.yml is the source of truth so re-run xcodegen generate after editing.
Runs ccusage's compiled binary directly (vendored in Vendor/ccusage/). Node is only needed for dev scripts, not for the built app.
Releasing
git tag vX.X.X && git push --tags # triggers .github/workflows/release.yml
That workflow builds, signs, notarizes, publishes a GitHub Release, and updates the Sparkle appcast (docs/appcast.xml) for auto-updates. Auto-updates only work once the repo is public (Sparkle needs public release URLs).
One-time setup for releases (Apple Developer account required) — set these GitHub secrets:
| Secret | Source |
|---|---|
DEVELOPER_ID_P12_BASE64 | Developer ID cert, exported as .p12, base64-encoded |
DEVELOPER_ID_P12_PASSWORD | That export's password |
TEAM_ID | e.g., Z4ZXD95E9B |
NOTARY_APPLE_ID | Apple ID email for notarization |
NOTARY_PASSWORD | App-specific password (appleid.apple.com) |
SPARKLE_PRIVATE_KEY | From Sparkle's generate_keys -x (public half already in project.yml) |
Local sanity check (no notarizing/zipping):
xcodegen generate
xcodebuild -project AIUsageMenuBar.xcodeproj -scheme AIUsageMenuBar -configuration Release build
Similar repos
Languages
Swift
76.1%
JavaScript
21.3%
Shell
2.6%