Claude Code usage report

    1

    Show what your Claude Code sessions cost, in tokens and dollars per day, week, project, model or session, from local transcripts with no network.

    Free

    1 useSecurity scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Claude Code usage report

    Example session with this skill installed

    /claude-code-usage-report

    In words: what did my Claude Code sessions cost in the last 7 days, by day, and which sessions cost the most?

    • Read your context and instructions
    • Compiled the claude code usage

    The skill ran its bundled script on your transcripts (here: a config folder with three synthetic sessions, so the numbers are small) and printed:

    Claude Code usage 2026-10-09 to 2026-10-15 (7 days), by day
    
    Day         Reqs  Input  Output  Cache wr  Cache rd  Fresh   Cost
    ----------  ----  -----  ------  --------  --------  -----  -----
    2026-10-09     0      0       0         0         0      0  $0.00
    2026-10-10     0      0       0         0         0      0  $0.00
    2026-10-11     0      0       0         0         0      0  $0.00
    2026-10-12     1   500k     80k      100k        2M   680k  $1.23
    2026-10-13     0      0       0         0         0      0  $0.00
    2026-10-14     2   130k     32k       30k      900k   192k  $1.51
    2026-10-15     3   380k    100k      150k      3.2M   630k  $4.93
    ----------  ----  -----  ------  --------  --------  -----  -----
    Total          6  1.01M    212k      280k      6.1M   1.5M  $7.67
    
    Most expensive sessions
       Cost  Fresh  Project  Last   Session
      -----  -----  -------  -----  ----------------------------------
      $4.10   635k  alpha    10-15  Synthetic session one (00000001)
      $2.34   187k  beta     10-15  Synthetic session two (00000002)
      $1.23   680k  alpha    10-12  Synthetic session three (00000003)
    
    Fresh = input + output + cache writes. Cache reads are cheap and listed apart.
    Dollars are API list prices. On a subscription plan nothing is billed per token: read them as how fast limits fill.
    

    Next steps the skill offers: report --by project (the same week per project), report --by model, today (cost, fresh tokens, cache reads, requests and the split by project for today), and --json on any of them. Nothing is changed: the script reads the transcripts and keeps an index of token counts under cost-guard/ in the config folder.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Claude Code usage report shows what your Claude Code sessions cost. Type /claude-code-usage-report and it prints a table of tokens and dollars for the last 7 days, or grouped by project, model, week or session, with a quick view of today and this week and a JSON option. It is free. Dollars are API list prices; on a subscription plan nothing is billed per token, so read them as how fast your limits fill.

    Everything stays on your machine. The skill runs a small bundled Node script (Node 18 or newer, no dependencies, MIT license) that reads the transcripts under your Claude config folder (CLAUDE_CONFIG_DIR, default ~/.claude) and adds up their usage records. It sends nothing anywhere and has no network code. To keep repeat runs fast it saves an index of token counts, session titles and project folders in a cost-guard folder inside your Claude config folder. The index holds no prompts, answers or tool output.

    The script is the bin, src, LICENSE and package.json of claude-cost-guard, unchanged from its release tag v1.0.1. For daily and weekly budgets with warnings inside Claude Code, see the full open-source tool: https://github.com/nrzz/claude-cost-guard

    Needs Node.js 18 or newer. Reads only the transcripts on this machine, under the Claude config folder: sessions on other machines or in the Claude apps are not in them. Dollars are API list prices, not what a subscription bills. Budgets and warnings are in the full tool (claude-cost-guard), not in this skill.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    Node.js 18 or newer

    Permissions

    Terminal / Shell
    Environment Variables
    Write Files
    Read Files

    File Scopes

    ~/.claude/projects/**
    ~/.claude/cost-guard/**

    Reads the transcripts under the Claude config folder to add up their usage records, and reads CLAUDE_CONFIG_DIR to find that folder. Writes only its index (token counts, session ids and titles, project folders; no prompts, answers or tool output) under cost-guard/ inside that folder. No network code, no browser.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Claude Code keeps each session as a transcript file under its config folder, and the usage numbers sit inside those files. The skill runs a small script that reads every transcript and adds the numbers up; asking the model cannot do that, because the model does not see the files or the prices.

    No. Everything runs on your machine. The script reads the transcripts under your Claude config folder, prints the report, and keeps an index of token counts in a cost-guard folder there. It has no network code and sends nothing anywhere.

    It is written for Claude Code: the skill pre-approves its two commands, so /claude-code-usage-report runs with no permission prompt. The script itself is plain Node.js, so any other agent, or you in a terminal, can run it by hand: node cost-guard/bin/claude-cost-guard.mjs report --days 7

    The prices are a table inside the script, the API list prices at release 1.0.1 of claude-cost-guard. Nothing is fetched. A new release of the skill updates the table. On a subscription plan nothing is billed per token, so read the dollars as how fast your limits fill.

    report --by project groups the week by project folder, and report --by model, --by week and --by session group it the other ways; today shows the split by project for today. The script reads all sessions in the config folder.

    New in LLM Ops