Back to browse
GitHub Repository

SDK to track cost-per-outcome for AI workflows

16 starsPython

Cost per Outcome for AI Workflows

by deborahjacob·Feb 24, 2026·4 points·4 comments

AI Analysis

●●SolidSolve My ProblemBig Brain

Cost attribution per event via OpenTelemetry, but Anthropic Workbench and Vercel do this already.

Strengths
  • Event-level cost tracking (not sampled) solves outcome-based pricing math that startups need today.
  • OpenTelemetry native means zero custom logging; auto-instrumentation for 50+ libraries reduces boilerplate.
  • Environment-variable config eliminates hardcoded API keys and cross-service coordination friction.
Weaknesses
  • Requires external OpenTelemetry Collector for production; adds infrastructure burden vs managed alternatives.
  • Positioned for AI platforms, but unclear how it compares to built-in cost tracking from Anthropic, OpenAI console.
Target Audience

AI platform builders, SaaS founders tracking per-customer AI costs

Similar To

Anthropic Workbench · Vercel AI cost tracking · LangSmith

Similar Projects