AI Agent Hub
Back to plugins
🖥️

dsh-token-stats

Client Updated 2026.08.24

Run the following command in DeepSeek Harness:

dsh plugin install QEDQCD/dsh-token-stats

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install QEDQCD/dsh-token-stats in DeepSeek Harness to install the plugin; the open-source repository is available at https://github.com/QEDQCD/dsh-token-stats

About this plugin

The built-in dsh-token-meter in DeepSeek Harness can tell you how many tokens a single turn consumed, but it will never answer how much you burned across the entire month. dsh-token-stats closes exactly that gap: it silently listens to every assistant message carrying a usage field, appends a lightweight record to a local JSONL log, and gives you an on-demand, cross-session aggregation report whenever you want one.

The core experience is straightforward. Ask the model about token usage for this month in any session and it calls the built-in token_report tool, returning a text table grouped by day, week, or month, filterable by date range and local or UTC timezone. Prefer the terminal? A single command prints the full report without spinning up a DSH runtime at all. Every record stores only token counts, session IDs, and timestamps, with no prompts, no API keys, and all data kept strictly on your machine.

It is built for developers who work daily with the DeepSeek Harness CLI and want a clear picture of their token spend. If you juggle multiple models, run long sessions, or simply want to know which days and which kinds of tasks cost the most, this plugin turns that answer from a manual log hunt into a single natural-language question.

Use Cases

  • Ask token usage in-chat and get an aggregated table instantly
  • Compare token consumption across different date ranges
  • Run a single command to view usage history without starting a DSH runtime

Best For

  • Developers working daily with the DeepSeek Harness CLI
  • Teams that need to track and control token costs
  • Individual users who want quantified insight into AI session usage