ccusage Reports Spend. SigRank Ranks Efficiency.
ccusage reads local agent logs and tells you what you spent. SigRank takes the same four token counts and asks how you compare with other operators. Different questions; here is when each is the right tool.
What ccusage is, as of 2026-11-11
ccusage is an MIT-licensed command-line tool that reads the usage data coding-agent CLIs leave on your machine and turns it into cost and token reports. Its README lists 19 supported sources, including Claude Code, Codex, OpenCode, Amp, Cline, GitHub Copilot CLI, Gemini CLI and Goose. Reports are daily, weekly, monthly and session, plus blocks (Claude Code 5-hour billing windows) and a beta statusline hook for Claude Code. You run it with npx ccusage@latest (or bunx / pnpm dlx), add --json for machine-readable output, and it reports cache-creation and cache-read tokens separately. The npm package ccusage was at version 20.0.49, published 2026-11-10.
Where ccusage is the better choice: it is free, needs no account, covers far more agent CLIs than most trackers, gives per-model dollar costs, and its blocks and statusline views help you watch a live 5-hour window. If your question is “what did I spend, on which model, on which day?”, ccusage answers it directly and SigRank does not try to.
Two different questions
ccusage is a ledger: how many tokens went in and out, and what that cost. SigRank asks a different question about the same four counts: relative to other operators, how much output and reuse did you get per unit of fresh input? It turns the counts into canon metrics and publishes them on a public board so you can see where you sit.
- Υ Yield is the headline: (cache_read × output) / input². It is large when cached context is reused heavily and output is high relative to fresh input.
- Leverage is cache reuse per fresh input; Velocity is output per fresh input (a ratio, not a time rate); SNR is the output share of fresh traffic.
- 10xDEV and Scale V (log10 of total tokens) round out the token-era board.
None of these says whether the work was good. They describe token-processing patterns, nothing more. The measurement guide walks through reading them, and the Yield calculator lets you try numbers by hand.
Use ccusage when, use SigRank when
- Use ccusage when you need dollars, a per-day or per-session breakdown, a 5-hour block view, or coverage of an agent CLI SigRank does not read.
- Use SigRank when you want to know whether your reuse-to-fresh-input pattern is high or low compared with other operators, track it over 7d / 30d / 90d windows, or publish a signed snapshot others can look at.
- Use both when you are tuning a workflow. A concrete loop: run npx ccusage@latest daily for the spend, then npx sigrank submit --dry-run to see the Yield, Leverage and Velocity of the same usage before you change prompts or caching habits, and again after. ccusage tells you what the change cost; SigRank tells you whether the pattern behind it moved.
Adding SigRank if you already run ccusage
You do not need to change how you use ccusage. The SigRank CLI is its own npm package (sigrank 1.0.49 as of 2026-11-11), and it bundles ccusage, tokscale and token-dashboard, so the log-reading layer is shared rather than replaced:
npx sigrank # inspect local usage in the terminal npx sigrank enroll # bind this device with a connect code from Settings npx sigrank submit --dry-run # inspect the payload; nothing is sent npx sigrank submit # publish a signed snapshot
Prefer not to install anything? The score calculator takes four token counts (input, output, cache creation, cache read), which you can copy from ccusage --json output, and returns a preview without an account. Ranked submissions do not need prompt text, responses or source code.
What the live board looks like
Fetched from the public SigRank analyst MCP on 2026-11-11. As of 2026-11-11 the 30-day public board had 7 public operators, so the field is small and a rank means little yet; the spread of Yield across operators says more than your position. The current count is on the live board and moves as people enroll; definitions are on the methodology page.
Side-by-side
| Question | ccusage | SigRank |
|---|---|---|
| Main question answered | What did I spend, and on what? | How does my reuse pattern compare with other operators? |
| Data read | Local usage data from 19 listed agent CLIs | Local usage via the SigRank CLI (bundles ccusage, tokscale and token-dashboard); four token pillars |
| Cost in dollars | Yes, per model (pricing data; --offline uses a cache) | Cost per 1M tokens appears on the board; not a billing ledger |
| Time views | daily, weekly, monthly, session, 5-hour blocks | 7d / 30d / 90d / all-time board windows |
| Efficiency metrics | Raw counts and cost | Υ Yield, SNR, Leverage, Velocity, 10xDEV, Scale V |
| Public ranking and profiles | No | Yes, after enroll and submit |
| Account required | No | Only to enroll and submit; the calculator needs none |
| Licence | MIT | See the repository LICENSE |
Frequently asked questions
- Does SigRank replace ccusage?
- No. They answer different questions. ccusage is a cost and token ledger across many agent CLIs; SigRank compares reuse-to-fresh-input patterns across operators. Keep ccusage for spend.
- Which ccusage numbers can I use in SigRank?
- The four token pillars: input, output, cache creation and cache read. Run ccusage with --json, copy those counts into the score calculator at /score, and you get a Yield preview without an account.
- Is a high Yield the same as a low ccusage bill?
- No. Yield is high when cached context is reused heavily relative to fresh input. That often, but not always, lines up with lower cost; ccusage is the tool to check the actual bill.
- How should I read a rank on a small board?
- As an early signal. The public board is small and changes as people enroll, so the spread of Yield across operators and your own trend over time tell you more than a single position.
- If SigRank depends on ccusage, why run ccusage separately?
- Because ccusage is the better tool for spend: per-model dollar costs, daily and monthly reports, and 5-hour block views. SigRank reports efficiency patterns, not a bill.
Ready to see your pattern?
Keep ccusage for the bill. Paste your four token counts into the calculator for a no-account preview, or enroll and submit a signed snapshot to appear on the board.
Sources and as-of date
Competitor facts were read from the sources below on 2026-11-11 (UTC); check them for anything newer. Field numbers come from the public SigRank analyst MCP on the same date.
Related: ccusage Alternatives · Yield Calculator · Methodology · How to measure AI coding efficiency · The Local Agent (MCP)
What the July 2026 field showed
In a frozen field of 1,642 operators, the highest-Yield operator ranked #697 by volume, and 92.4% of the median operator’s tokens were cache reads. Volume and efficiency are different questions. The findings: Volume Isn’t Yield · The Operator Explains 95% of the Difference · The Three Modes Behind the Board.
Already using ccusage? Keep it. Run npx sigrank on the same local logs to see your Yield next to the field. Token counts only, never prompts.