Show HN: AgentTrace–Observability and runtime self-healing engine for AI agents

Category: observability

Tags: ai-agents, observability, self-healing, llm, tracing

Score: 5.8/10 (Innovation: 5, Technical: 6, Documentation: 6, Utility: 6)

AgentTrace is an observability SDK and dashboard for AI agent pipelines that monitors tool calls, visualizes latency, and automatically repairs malformed LLM tool arguments at runtime to prevent crashes. It combines tracing with a self-healing layer powered by Groq for payload repair, filling a niche in production-grade agent reliability.

Target audience: AI engineers and backend developers building LLM agent pipelines

Repository: https://github.com/mohitkumar188/AgentTrace · Python · 3 stars

View on Hacker News