Context Rot Detection

Example Use Cases

• 1

  Real-time visibility into their cognitive state and context health.

• 2

  Actionable recommendations to mitigate context rot and maintain decision quality.

• 3

  That require tracking of health history and service-wide utilization statistics.

Description

This MCP server monitors and scores the health of AI agents' context windows to detect context rot, a common failure mode where performance degrades as context length increases. It offers model-specific degradation curves, lost-in-the-middle risk scoring, session fatigue analysis, and actionable recovery recommendations. The service supports over 15 curated models and auto-resolves HuggingFace models, storing health history and usage statistics in SQLite for ongoing tracking and optimization.