Self-hosted LLM API cost monitor. Per-request tracking for OpenAI, Anthropic, Google, and more.
Trough tracks cost per request, per model, per service. It supports OpenAI, Anthropic, Google, and other LLM providers. Anomaly detection and spend alerts are included in Pro.
curl -fsSL https://stockyard.dev/trough/install.sh | sh
LLM API costs are unpredictable by design. A prompt that costs $0.002 today could cost $0.02 tomorrow if your traffic pattern changes or the provider adjusts pricing. Helicone was the leading cost tracking tool until Mintlify acquired it in March 2026 and put it in maintenance mode. Langfuse starts at $59 per month. Self-hosted cost tracking sits between your application and the LLM provider, logging every request with its token count and cost. You see spend in real time without sending your prompts through a third party.
A self-hosted LLM cost tracker should proxy requests transparently — your application changes one URL and everything else stays the same. It should record token counts, latency, and calculated cost per request. It should aggregate spend by model, by day, and by upstream provider. And it should alert when spend exceeds a threshold so a runaway loop does not generate a surprise bill.
Trough proxies HTTP requests to any upstream API, records request and response sizes, calculates cost based on configurable rules, and aggregates spend over time. The dashboard shows total spend, spend by upstream, and request volume. The API exposes spend data, request history, and cost rules. CSV export lets you pull spend data into spreadsheets for reporting. Spend alerts notify you when costs cross a configured threshold.
LLM API costs are unpredictable. A prompt change can double your bill overnight. Self-hosted cost tracking gives you per-request visibility without sending your prompts to another third-party service.
Replaces: Helicone (acquired), Langfuse ($59/mo)
1 service, 10,000 requests/mo
Unlimited usage, full features
Single binary. Free to start. $2.99/mo for Pro.