Are you the author? Sign in to claim
Native macOS app for tracking Claude Code usage — tokens, cost, rate-limit pacing, per-project breakdowns. SwiftUI + Swi
Know what Claude Code is costing you — and how close you are to your limits — right from your Mac's menu bar.
Pacer is a free, open-source macOS app that keeps an eye on your Claude Code usage: the tokens you're burning, what they cost, how close you are to your 5-hour and weekly rate limits, and where the spend is going — by project, by model, by day. It sits quietly in your menu bar, keeps itself up to date, and keeps all of your data on your own Mac.
Menu bar & click-down popover |
Home-screen & Notification Center widgets |
Share your 5-hour & 7-day burn charts as an image
Models — share, trend, and per-model table; group by model or class
Projects & Collections — group projects, scope totals to a collection
History — lifetime totals, six-month heatmap, monthly spend
Heads up — early days. Pacer is pre-1.0 and under active development. It works and it's useful today, but expect the occasional rough edge, and your saved history may need to be rebuilt between 0.x versions.
Pacer-x.y.z.dmg file under "Assets").That's all. Pacer keeps itself up to date automatically: when a new version ships, it offers to install it for you — no re-downloading, no reinstalling.
You'll need macOS 15 (Sequoia) or newer, on either Apple Silicon or Intel.
You've got options for keeping an eye on Claude Code usage — Claude Code's own
/usage, the popular ccusage CLI, and
several menu-bar apps — the closest being Token Pacer,
a paid, closed-source namesake, plus Claude God and
ccseva. They're good tools; here's
the honest lay of the land.
| Pacer | Token Pacer | Claude God | ccseva | ccusage | CC /usage | |
|---|---|---|---|---|---|---|
| Form factor | Menu-bar app | Menu-bar app | Menu-bar app | Menu-bar app | CLI | In-terminal |
| Native macOS (not Electron) | ✓ | ✓ | ✓ | – (Electron) | – | – |
| Always-on, glanceable | ✓ | ✓ | ✓ | ✓ | – | – |
| Live limit % from Anthropic's API | ✓ | ✦ | ✓ | ✦ | ✦ | ✓ |
| Pace vs. ideal-burn line | ✓ | – | – | – | – | – |
| End-of-day spend projection | ✓ | ✓ | ✓ | ✓ | – | – |
| Activity heatmap (6 months) | ✓ | – | – | – | – | – |
| Per-project & per-model breakdown | ✓ | ✓ | ✓ | ✓ | ✓ | – |
| Beyond Claude Code (Codex, opencode, …) | – | ✓ | – | – | – | – |
| Home-screen / Notification Center widgets | ✓ | – | ✓ | – | – | – |
| Threshold / budget notifications | ✓ | ✓ | ✓ | ✓ | – | – |
| CSV export | ✓ | ✓ | ✓ | – | ✓ | – |
| ROI: cost vs. git commits | – | – | ✓ | – | – | – |
| Claude Code plugin marketplace | – | – | ✓ | – | – | – |
| Price | Free | $29 one-time | Free | Free | Free | Included |
| Free & open source | ✓ | – | ✓ | ✓ | ✓ | – |
✦ estimates the windows from your local JSONL logs rather than reading Anthropic's usage API. Token Pacer is a paid, closed-source app, so its column is read from its public site rather than its source — and some of its features (CSV export, custom model pricing) are Pro-tier. Best-effort as of June 2026 — these tools all move fast, so corrections are welcome via an issue or PR.
The gist:
/usage — Claude Code's built-in view is a snapshot you ask for in
one terminal; Pacer is the always-on, zoomed-out companion that remembers every
session. The speedometer in one car vs. the dashboard that logs every trip.ccusage — a great CLI for scriptable numbers (Pacer's tests even
cross-check their scanner against it); Pacer is the GUI you glance at instead of
a command you re-run, and it reads your actual limit % from Anthropic rather
than estimating from logs.Pacer is local-first. It reads only the files Claude Code already writes to your own Mac, and it sends nothing about your usage anywhere:
~/.claude/projects/*.jsonl — your session token usage, read on your machine.The only network connections Pacer makes are:
api.anthropic.com — to check your rate-limit windows (~12 small requests an
hour while it's running).github.com — to check for, download, and install app updates (at launch,
then every 6 hours; installs happen in the background when automatic updates
are on).No analytics, no telemetry, no third parties. Your data stays on your Mac in
~/Library/Group Containers/…/pacer.sqlite, and it persists across app updates.
Pacer is SwiftUI + SwiftData + Charts, signed with Developer ID and notarized,
auto-updating via Sparkle. Contributions welcome —
see CONTRIBUTING.md.
| Component | Role |
|---|---|
Pacer.app | Main UI + menu-bar item. Data collection (FSEvents JSONL scan + OAuth poll) runs in-process inside this binary — there is no separate daemon. |
PacerWidgets | WidgetKit extension — reads the shared App Group store directly. |
PacerCore | Shared Swift package — parsers, models, scan coordinator, recomputers. |
For everyday use, the released DMG is what you want — this section is only for hacking on Pacer.
Requirements: macOS 15 SDK (Xcode 16+) and xcodegen (brew install xcodegen).
The Xcode project is generated from project.yml — never edit the .xcodeproj
directly.
make verify # unsigned compile-only check (no Apple account needed)
make test # PacerCore unit + ground-truth tests
make install # signed + notarized build → /Applications/Pacer.app
make screenshots # regenerate the README screenshots (see docs/screenshots.md)
make help # everything else
make verify and make test need no signing setup and are all that CI runs.
To build a runnable app you need your own Apple Developer account — point the
install at it with PACER_SIGN_IDENTITY="Developer ID Application: <Name> (<TEAMID>)" make install.
See CONTRIBUTING → "Building and running it yourself"
for the full story (and why an App Group ties signing to your Team ID).
AGENTS.md is the deep architectural guide (performance invariants,
the SwiftData schema, the recomputer pattern); docs/ covers the design,
perf-tuning, screenshot generation, and release process.
Pushing a vX.Y.Z tag triggers
.github/workflows/release.yml, which builds,
signs, notarizes, packages a DMG, signs the Sparkle update, publishes the GitHub
Release, and updates appcast.xml on the gh-pages branch. See
docs/releasing.md for the secrets and the cut-a-release
checklist.
MIT. Pacer reads data from Anthropic's Claude Code product but is not affiliated with or endorsed by Anthropic.
1000+ skills curated from Anthropic, Vercel, Stripe, and other engineering teams
Agent harness performance optimization with skills, instincts, memory, and security
Design enforcement with memory — keeps your UI consistent across a project
Detects 37 AI writing patterns and rewrites text with human rhythm across 5 voice profiles