Show HN: Ark – AI agent runtime in Go that tracks cost per decision step

Category: ai-ml

Tags: ai-agents, llm-runtime, context-optimization

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

Ark is an AI agent runtime written in Go that dynamically manages LLM context by loading only relevant tools per task, reducing token overhead by ~99%. It learns from each execution to rank tools based on success rate, cost, and latency, and provides per-decision cost attribution, making it an economically rational runtime for adaptive agents.

Target audience: backend devs, ai engineers, devops

Repository: https://github.com/atripati/ark · Go · NOASSERTION · 7 stars

View on Hacker News