Moth - Project-local bug-fix analysis and verified fix memory.

Hi Product Hunt! 👋 I'm the creator of Moth. Why I built it: I was incredibly frustrated by "agent amnesia". Every time I used an AI coding agent (like Claude Code or Cursor), it would face an error, blind-guess a fix from unredacted logs, and then repeat the exact same mistake a week later on a different session. The approach: I didn't want to build a heavy Vector DB or send telemetry to a cloud service. I wanted a simple, local, and standard protocol. So, I built Moth. It's an open-source MCP (Model Context Protocol) server. Instead of just pasting logs, Moth exposes two tools to your agent: analyze_error: Automatically redacts secrets, detects the framework stack, and checks past memory before the agent guesses a fix. remember_fix_result: If a fix works and passes a verification command (like npm run build), it records it locally in a read-only .moth/fix-memory.jsonl file in your repo. It's incredibly explicitly designed to keep the AI workflow structured and safe. I’d love to hear your feedback on this local-first approach to AI agent memory! Let me know if you run into any issues.