See what your agents cost and how you work
Stats shows what your sessions and AI helpers cost, how much time they took, and how you use Yardmaster, for any time range.
Stats answers questions like "what did my agents cost this week?", "which project uses the most?" and "how many hours did I work?". It reads what Yardmaster already records on this computer and shows it as cards, charts and lists.
Open Stats#
Click Stats in the sidebar.
Pick a tab under Tabs, for example Overview or Spend.
Pick a period under Time Range: Today, Yesterday, 7 Days, 30 Days, 90 Days or All Time.
The refresh button next to the Stats title reloads the numbers.
What the dollar amounts mean#
Where the numbers come from. For Claude sessions, Yardmaster reads the transcripts Claude Code saves on this computer, one entry per request, and prices each one. Other engines (Codex and the rest) count what that engine reports.
What is included. Total Cost covers your coding sessions, including sessions you have since deleted. The Spend tab adds the background helpers that call an AI model, such as session titles and summaries.
The limits. A session whose transcript is no longer on this computer (for example one run over SSH) cannot be priced from its transcript, so its older cost is left out. Its new requests still count, from the totals Claude Code reports. A model with no known price shows as such under Accuracy (see below). Treat the numbers as a close estimate, not an accounting record.
The tabs#
Overview: headline cards (Total Cost, Total Tokens, LLM Time, Active Time, Sessions, Your Messages, Automated Actions, Typing Speed), then By Project and Top Sessions.
Spend: Where your AI money goes, one list, biggest first: your coding sessions plus every background feature that called a paid model. Click a feature to see its individual charges. Spend by engine splits the coding sessions by engine.
Reliability: time lost to network, sign-in and rate-limit problems, retries, API errors, accounts at their limit, problem tools and helper-agent spend. Claude sessions only.
Cost by cause: why Claude sessions cost what they did: how large sessions grew, Cold cache rewrites (a large session picked up again after its cache expired, so everything was sent again), sub-agents and the Most expensive sessions.
Feature Usage: how often each Yardmaster feature ran, and its cost if it calls a paid model.
Shortcut Efficiency and Hotkey Usage: how often you use the mouse or the keyboard for common actions, and which shortcuts you press most.
Trends: sessions and cost per day, your streak and busiest day.
Usage: how fast you use up your Claude plan's limits; see Know before you hit a usage limit.
Records: your all-time personal bests.
Settings Search: what you searched for in Settings and how high the result you picked was listed.
The Time Range list shows on every tab except Trends, Records and Settings Search. Records and Settings Search always cover everything Yardmaster has recorded. Trends has no list of its own: it shows the period you last picked on another tab, and there All Time means the last year. Usage shows the list but ignores it.
Break a number down#
Click the Total Cost, Total Tokens, LLM Time or Sessions card on Overview to open the Usage breakdown for the chosen range. For one session, click the cost in the session's header.
By model: cost, requests and every kind of token per model.
Over time: cost or tokens per hour, day or week.
Where it went: main sessions versus sub-agents and asides, and each project's models.
Accuracy: how much of the figure is exact, any model with no price, and a check against Claude Code's own totals.
Why your older numbers changed#
Earlier versions added up Claude Code's running totals on every reply, so long sessions showed far too much cost and LLM Time. Each request is now counted once. The first time a newer version starts, it re-reads the transcripts of your existing sessions in the background and corrects their figures. While it runs, Overview shows "Correcting usage history from transcripts" with a count.
Sessions with no transcript. If a session's transcript is gone from this computer, its earlier cost, tokens and LLM Time are taken out of every total (the session still counts as a session), and Overview says how many.
When the correction waits. If Yardmaster cannot read the transcript folder, the folder is on a drive that is not connected, or it holds far fewer transcripts than expected, it changes nothing and says the correction is on hold. It checks again every 15 minutes.
Active Time counts time Yardmaster was in front or a session was working, not time the app sat idle.
Records lists bests in Peak Parallelism, Single-Session Feats, Efficiency, Lifetime, Streaks, Daily Bests and Leverage. A record you have not reached shows locked. A best tied to one session or day opens it when you click it. Confetti plays on the tab when you beat a record.
Reliability keeps its events for 90 days and samples account limits every 15 minutes.
Known subtotal. If you use sessions billed through OpenRouter, the cost card reads "Known subtotal", because some charges are not known yet.
Hear about a new record#
Two switches in Settings > Notifications, both off at first:
The inbox card is kept rare on purpose: only clear jumps on milestone records, and at most a couple a day.
Does Stats leave my computer?#
The Stats page itself is built on this computer. While you are signed in, though, Yardmaster sends numbers to the Yardmaster cloud, tied to your account: that day's totals (sessions, messages, tokens, LLM time, active time, AI spend), your all-time totals and personal bests, a random id for this copy of Yardmaster, the app version, your operating system and your Claude plan name. It sends numbers only: no project names, session titles or anything you typed. Privacy and your data lists everything that leaves your computer.
Related
- Keep long sessions affordableEvery reply re-reads the whole conversation. Yardmaster warns you when a session gets large, holds costly automatic replies, and can continue in a fresh session.
- Get a daily AI spend card in your InboxTwo ways to get a card each morning about yesterday's AI spend: the built-in Daily spend summary or the Daily Spend Report plugin.