ccusage does detailed token and cost reporting well, especially for developers who already live in a shell. People usually search for a ccusage alternative because they want an always-visible desktop view, support for a wider set of tools, or a UI that does not require running a command. Others still want CLI output but need project attribution, more data sources, or a local dashboard. The right replacement depends on whether usage history or live attention is the actual problem.
1. codeburn
Best for: the closest practical replacement for ccusage on its strongest dimension
codeburn reads local records from a broad set of coding tools and turns them into cost, token, model, project, and task views. Its 31-tool scope is wider than most focused trackers. The product is a strong replacement when the reason for switching is fragmented spend data rather than waiting for an agent to finish.
Strengths
Broad tool coverage, useful project and task attribution, local processing, MIT source, and a quick npx entry point.
Trade-offs
It is centered on analysis after activity is recorded. Ambient session state and your-turn alerts are not found in the reviewed public documentation.
- License
- MIT
- Price
- Free
- Platforms
- Terminal and local web UI; check current package requirements
- Official source
- codeburn
2. Tokscale
Best for: usage analysis, not a turn-state companion
Tokscale is a fast CLI and visualization dashboard that reads many coding-agent stores and calculates token use and cost. It covers far more sources than a five-provider status app and can create a public profile when the user explicitly submits data. It suits developers who enjoy terminal reports and want a single usage ledger across a long tool list.
Strengths
Wide agent support, CLI speed, cost calculations, MIT source, and optional shareable profiles.
Trade-offs
Its center of gravity is usage analytics. Public profile submission is optional but separate from a strictly local-only workflow, and live your-turn state is not the product focus.
- License
- MIT
- Price
- Free
- Platforms
- macOS, Linux, Windows where supported by its runtime
- Official source
- Tokscale
3. claude-usage
Best for: historical session and usage dashboard
claude-usage turns Claude Code JSONL files into terminal summaries and a local dashboard. It is dependency-light, works across the major desktop operating systems, and exposes sessions, projects, models, token counts, and cost estimates. It is a practical choice when the reader wants to inspect history rather than monitor several agents at once.
Strengths
Clear local data path, useful dashboard and CLI views, broad installation choices, and cross-platform support.
Trade-offs
It only covers Claude Code and relies on scanning recorded sessions. It does not provide a menu-bar state model or documented your-turn notifications.
- License
- MIT
- Price
- Free
- Platforms
- macOS, Linux, Windows
- Official source
- claude-usage
4. Agent Island
Best for: live for claude, codex, gemini, and grok; cursor usage only
Agent Island is a desktop status companion rather than a report you invoke. It keeps supported agent state, quota windows, local cost estimates, and report cards in one place. Claude, Codex, Gemini, and Grok contribute live state; Cursor contributes billing-cycle usage and plan data only.
Strengths
Cross-platform ambient status, a shared five-provider Usage view, inspectable MIT source, and no Agent Island account or telemetry.
Trade-offs
Cursor has no live session state. Gemini reports working or idle without turn-boundary alerts. Cost figures are estimates from public per-token prices, not billing records, and the project is smaller than the largest tools in this list.
- License
- MIT
- Price
- Free
- Platforms
- macOS, Windows
- Official source
- Agent Island
5. TokenTracker
Best for: usage activity and hooks
TokenTracker combines a local usage dashboard with native tray apps, widgets, achievements, and a desktop pet. It documents 29 coding tools and runs on macOS, Windows, and Linux. The extra presentation layers make it a better fit than a sparse CLI for people who want usage to remain visible throughout the day.
Strengths
Very broad tool coverage, three desktop platforms, local-first setup, dashboard depth, and native widgets.
Trade-offs
Its richest surfaces focus on token activity, cost, and gamification. Readers seeking a restrained working, stalled, and your-turn indicator may find the scope larger than necessary.
- License
- MIT
- Price
- Free
- Platforms
- macOS, Windows, Linux
- Official source
- TokenTracker
Comparison table
| Alternative | License / price | Platforms | Agents | Live state | Your-turn alerts | Cost | Data | Install |
|---|---|---|---|---|---|---|---|---|
| codeburn | MIT; Free | Terminal and local web UI; check current package requirements | 31 tools and agents documented | Historical usage and task analysis | Not found in reviewed public documentation | Local token and cost tracking | Local | npx |
| Tokscale | MIT; Free | macOS, Linux, Windows where supported by its runtime | Many coding agents including Claude, Codex, Gemini, Cursor, and Grok | Usage analysis, not a turn-state companion | Not found in reviewed public documentation | Token and cost calculation | Local by default; optional public profile submission is documented | bunx / package instructions |
| claude-usage | MIT; Free | macOS, Linux, Windows | Claude Code | Historical session and usage dashboard | Not found in reviewed public documentation | Local cost estimates | Local JSONL into a local SQLite database | Homebrew, uv, pipx, clone, or Docker |
| Agent Island | MIT; Free | macOS, Windows | Claude, Codex, Gemini, Grok, Cursor | Live for Claude, Codex, Gemini, and Grok; Cursor usage only | Your-turn alerts where the provider exposes that boundary | Local estimates | Local; no account or telemetry | GitHub Releases |
| TokenTracker | MIT; Free | macOS, Windows, Linux | 29 AI coding tools documented | Usage activity and hooks | Not documented as a shared your-turn state layer | Token and cost dashboard | Local-first; no cloud account or API keys claimed | npx, native downloads, or Homebrew |
How to choose
Start with the interface you will keep open. An ambient menu-bar app reduces checking; a CLI gives precise filters on request. Decide whether one provider deserves deep treatment or several agents need one view. Inspect the license and data path, confirm operating-system support, and separate local estimates from raw token counts or billing records.
- Do you want ambient status or an on-demand report?
- Is one agent enough, or must several providers share one view?
- Do you need inspectable source and a strictly local data path?
- Must the same workflow run on Windows as well as macOS?
- Are estimates useful, or do you need a specific raw-count export?
Migrating from ccusage to Agent Island
Install Agent Island from the latest GitHub Release, then leave ccusage installed during the trial. The tools can read their own supported local sources without importing a ccusage database. Replace the habit of running a block or daily report with Agent Island's Usage and report-card views, and use its status surface for return-to-work timing. Keep ccusage when you still need its exact terminal filters or scriptable output; Agent Island does not reproduce every CLI report.
Want the head-to-head? Read Agent Island vs ccusage
FAQ
Is there a free alternative to ccusage?
Yes. The open-source entries here are free. Pick by job: terminal reporting, quota pacing, or ambient desktop status.
What is the best ccusage alternative for macOS?
Choose a native menu-bar or notch app for ambient visibility. Choose a CLI or dashboard for detailed historical queries.
Can I run ccusage and Agent Island together?
Yes. Disable duplicate notifications, compare the same sessions, and remove one only after its unique features are unnecessary.
Does ccusage support Codex, Gemini, or Grok?
Check the linked official documentation for the current provider list. This page does not infer undocumented support.
Which alternative works on Windows?
Agent Island and the cross-platform entries marked in the table have documented Windows routes. Mac-native apps do not.
Do these tools upload prompts or source code?
Data paths differ. Read the linked source and current privacy terms before granting credentials or optional sharing permissions.