dsh-spend

For developers who frequently call multiple LLM services, usage and cost data are often scattered across different provider consoles with inconsistent metrics and billing models, making real-time cost tracking a real challenge. dsh-spend is a one-stop usage and billing monitor built for DeepSeek Harness users. It presents a persistent floating capsule in the bottom-right corner of the Web UI—hover to preview cost and token summaries, click to expand a four-tab panel covering Overview, Today, Performance, and Call Details to analyze every call from dimensions such as provider, model, hour, day, session, and working directory.

The plugin ships with a built-in knowledge base of official pricing for 17 providers and 131 models, enabling zero-config auto-detection of subscription (Code) and pay-as-you-go (Token) billing plans. For Code plans, it directly reads provider APIs to display real quota usage percentages and reset times; for Token plans, it estimates costs based on official unit prices, with support for DeepSeek peak/off-peak pricing, real-time $ / ¥ currency switching, and workspace filtering. Call details can be exported as CSV / JSON, and abnormally expensive calls are highlighted in red to help spot cost anomalies quickly.

Whether you are an individual developer aiming to control monthly AI spending or a team lead needing per-project cost accounting across multiple model providers, dsh-spend delivers a zero-config, real-time, transparent view of token usage and expenses, making every token consumption clearly visible.