← Index

GATEAgents & automation

Jev 1.13 MCP Failure Response

Project photo 1

This project investigates how Jev 1.13, an AI agent, responds to deterministic MCP failures and if successful recovery evidence alters its decision-making process.

I tested how @typesafeai’s Jev 1.13 responds to deterministic MCP failures—and whether successful recovery evidence changes its decision. I used MCP Failure Lab through @OpenRouterAI: 6 controlled scenarios, 20 runs each. Results below. Full breakdown in the thread 🧵 #MCP #AIAgents