Show HN: We stress-tested LangGraph with an evolutionary fuzzer

Category: ai-ml

Tags: ai-agent-testing, fuzzing, quality-diversity, regression-testing, mcp, llm-safety

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

Life Forge is a behavioral regression testing framework for AI agents that simulates tool side effects, uses MAP-Elites quality-diversity search to find failure scenarios, and converts them into deterministic replayable regression tests. It stands out by judging actions on simulated world state rather than LLM-judged text, and supports MCP-native domain compilation plus CI gates.

Target audience: AI engineers, ML platform teams, and security engineers building tool-using agents

Repository: https://github.com/zariffromlatif/life-forge · Python · MIT · 5 stars

View on Hacker News