AI Agent Hub
Back to plugins
🧠

dsh-cost-ledger

Memory Updated 2026.08.15

Run the following command in DeepSeek Harness:

dsh plugin install suimi8/dsh-cost-ledger

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

Install the plugin in DeepSeek Harness by running dsh plugin install suimi8/dsh-cost-ledger; the source repository is https://github.com/suimi8/dsh-cost-ledger . Restart dsh web after installation to activate.

About this plugin

When you drive multi-session agent workflows with DeepSeek Harness, every LLM call quietly burns tokens yet native cost tracking is often missing or scattered across logs. dsh-cost-ledger captures the token usage of every model call (including retries) and persists it to a local SQLite database, giving you a cross-session, queryable cost ledger that actually sticks.

By subscribing to the host llm/stream waterfall, the plugin records a timestamp, session id, model, input/output/cache-read/cache-write token counts, and a fee computed from built-in DeepSeek official pricing with fully cache-aware billing rules. You can override or extend rates for any model via config. Alongside the passive logging, the plugin exposes three agent tools — record_cost, query_cost, and set_budget — so you can pull spend summaries with per-model breakdowns by day, week, or month, and create or adjust budget thresholds scoped by daily, model:, or project: labels, keeping API spend always in check.

Ideal for developers and teams who rely on DeepSeek Harness for agent development, batch jobs, or model comparison experiments and need concrete visibility into token spend broken down by project or model, plus budget guardrails on top. The plugin makes zero outbound network calls and stores no prompt or completion content — only aggregate token counts and metadata — and works right after install with no extra wiring.

Use Cases

  • Auto-capture every LLM call's token usage and computed cost across multi-session agent workflows
  • Break down API spend by model or project, set daily/weekly/monthly budget thresholds, and check overage status
  • Persist cost data across sessions for post-hoc auditing and real cost comparison between models

Best For

  • Developers and engineering teams building agent applications on DeepSeek Harness
  • Tech leads who need visibility into and control over LLM API token spend
  • Engineers running batch inference jobs or multi-model A/B experiments who need cost breakdowns