InsAIts – Open-source monitoring to detect AI-to-AI hallucinations

Summary: InsAIts is an open-source tool designed to detect hallucinations and anomalies in multi-agent AI systems by monitoring AI-to-AI interactions. It identifies contradictions, fabricated citations, confidence decay, and grounding issues to prevent error propagation across agents.

What it does

InsAIts uses five hallucination detection subsystems and six anomaly detectors to monitor AI agents communicating, catching issues like cross-agent contradictions, phantom citations, document grounding, and confidence decay. It integrates with LangChain, CrewAI, LangGraph, and exports to Slack and Notion, running entirely locally with privacy-first design.

Who it's for

It targets developers and teams building multi-agent AI systems who need to track and prevent error propagation between AI agents.

Why it matters

It solves the problem of silent error amplification in AI pipelines by providing a monitoring layer that detects and traces hallucinations before they affect outputs.