Getting started
Your first report
Run it with no arguments. There is nothing to configure, no account, and no first-run wizard.
$ actualisOn a machine with real history this takes a few seconds and prints a report in
sections. If it is slow, narrow the window — --days 7 reads far
fewer files, because a file untouched since the cutoff cannot hold a record after
it.
What you get
| Section | Answers |
|---|---|
| FLEET | Window, how much was read, total cost, burn rate per active day. |
| TOKENS | Input, output, and cache split by write duration and read. |
| BY MODEL | Which models did the work, and what each cost. |
| BY PROJECT | Where the money went. |
| CACHE EFFICIENCY | Share of context served from cache, and what caching saved. |
| BY TICKET | Cost attributed to an issue, via the branch a message was written on. |
| SHELL AUDIT | Commands that ran, which are worth a look, and which carried credentials. |
| COACH | Findings worth acting on, benchmarked against your own history. |
Reading the numbers honestly
Three things are easy to misread, and the tool tries to say so in place:
- Cost is notional. It is what the recorded token usage would cost at provider list prices. On a Pro or Max subscription your actual outlay is the flat fee — read the figure as consumption, not as a bill.
- Flags mean "worth looking at", not "wrong". Most
rm -rfcalls are a build directory. - Rates use active days, not calendar days. One stale session from six months ago would otherwise stretch the window and quietly divide your weekly rate by five.
- Records are not messages. A transcript repeats the same assistant record while a response streams, so one billable message appears many times. Each is counted once, keyed on its message id, and the report prints how many repeats it collapsed. Versions before 0.1.1 billed every record and overstated a real corpus by 2.13×; if you have a figure from 0.1.0, re-run it.
Ask it how it got there
No figure is presented that cannot be traced back to the records behind it.
$ actualis --explain cost # the arithmetic, the rate, what was excluded $ actualis --why AF004 # why one finding fired, with your numbers