Balagan Agent – Chaos engineering for AI agents before production

Summary: Balagan Agent applies chaos engineering to AI agents by injecting controlled failures like timeouts, rate limits, malformed outputs, and latency spikes. It measures mean time to recovery (MTTR) and reliability scores, providing reports in terminal, JSON, Markdown, or HTML formats.

What it does

It wraps AI agents to simulate faults such as hallucinations, context corruption, and budget exhaustion, then tracks recovery metrics. It includes wrappers for CrewAI, Claude Agent SDK, and examples for AutoGen and LangChain.

Who it's for

Developers building AI agents intended for production environments who need to test resilience against real-world failures.

Why it matters

It identifies and quantifies failure modes before deployment, helping improve agent reliability and reduce silent failures in production.