inferock-bench
Local auditing of LLM costs, usage, and failed-call receipts
About inferock-bench
A locally running LLM cost-tracking proxy supporting OpenAI, Anthropic, Gemini Developer API, and fixed OpenRouter endpoints. It measures calls routed through localhost, writes local event records, and renders receipts using the included @inferock/measure grading code, helping audit AI/LLM bills, measure Claude or GPT token usage, and check whether failed API calls are billed.
Best for
Best use cases
Audit AI/LLM bills
Measure Claude or GPT token usage
Check whether failed API calls are billed
Record and render API call receipts locally
Features
Local proxy routing
Routes calls through localhost and writes them to a local event log.
Billing and usage audit
For auditing AI/LLM bills and measuring Claude or GPT token usage.
Failed-call billing check
Helps answer whether failed API calls are billed.
Receipt rendering
Renders receipts using the included @inferock/measure grading code.
Pros
- Covers OpenAI, Anthropic, Gemini Developer API, and fixed OpenRouter endpoints.
- Provider keys are attached only to provider requests; receipts are generated locally.
Cons
- Cannot audit traffic that bypasses the local proxy.
- Cannot limit provider spending for unobserved calls or explain monthly bills without a matching invoice.
Pricing Plans
Free
- Local benchmark use
* Pricing information is for reference only. Please refer to the latest announcement on inferock-bench's official website.