Use the local AI usage,
Organized into a submission package.

I want to do a cost checkup, but I don’t know how to prepare the materials?
Use TuneMeter to collect dosage, follow the manual to export and check, and then submit it to Tunenomics for verification and analysis.

01 Select range Only collect the sources and dates you selected
02 Look at desensitization Check submission package and known data gaps
03 Do a checkup Submit on your own initiative and return to the original task to receive the report

Follow the manual Get TuneMeter (new window)

v 3.3.5 Pre-release version · If you already have a bill or usage meter, you can also directly Enter checkup

Follow along / An operation manual

From the usage of the machine, to an analysis report.

Export first, then submit. Divide it into two chapters and do one step at a time.
The left and right direction keys can also turn pages, and the content will not automatically jump away.

FIELD GUIDE / EXPORT 01 / 08

Download CLI and unzip it completely

Open the official release page and select CLI ZIP. Unzip the entire folder and run it in the unzipped directory; it is not necessary to install the Agent Skill first.

Prepare tunemeter.pyz and package instructions.

Operation instructions Not reading your device

TuneMeter CLI

v 3.3.5 · Pre-release version

Download CLI ZIP (new window)Version notes and verification files This is the program package; the "Tuninomics submission package.zip" generated after running is the checkup material.

A / Local preparation One page at a time, at your own pace.

Page 1 / 8 · Get Tools

Nine types of known format adapters have been implemented, and compatibility is affected by the client format; the direct verification environment is Windows 11 / Python 3.12.14, which cannot be regarded as full platform or version certification.View support range

See the boundaries first

Usage records are the starting point for analysis.

Complete work instructions(new window)

Implementation scope

Collection, summary and default desensitization

Public v 3.3.5 Registered nine categories of known format adapters. Fixed fictional format having runtime validation on Windows 11 / Python 3.12.14; this does not mean it is compatible with all real client versions or platforms.

Item-by-item support matrix(new window)

Evidence of expenses

The collection tool does not generate actual deductions

This tool does not convert Token prices or calculate subscription payments. Incomplete or uncollected data cannot be recorded as zero. Tunenomics will check the measurement basis based on the selected materials and distinguish between observations, estimates and actual cost evidence in the report.

Default desensitization and sharing boundaries(new window)

What should I do if I encounter a gap or blocked?

Let’s look at completion_note, coverage_gaps, warnings and delivery first. exit 4 indicates a gapped result, which can usually still be reviewed; when delivery is blocked, the package is not officially submitted. Do not manually change the total to bypass the check, and do not upload the entire local diagnostic directory.

Reading status and exit code description(new window)

From tools to delivery

You control the material and leave the analysis to Tunenomics.

Let’s look at the delivery report first

01 / Ready

Submit reviewed materials

TuneMeter submission package, or existing PDF, CSV, Excel. Explain the source, date and problem you want to solve, and submit it proactively after logging in.

02 / Want to know first

See clearly what the data can explain

Collection package reflects observed usage. Actual expenses still require corresponding evidence, and subscription payments cannot be calculated based on Token alone; the report will explain the measurement basis and limitations.

Communicate the scope of analysis first

Checkup supports PDF / CSV / Excel / ZIP, a single file does not exceed 20 MB. Please review the desensitization instructions before submitting your ZIP. TuneMeter does not read account credentials and does not automatically upload; when the report is opened and given to the cloud agent for reading, the selected content may still be passed to the model service.