Relays detailed error logs from other MCP servers to LLMs, enabling more intelligent error recovery and reducing token waste.
When LLM calls to MCP servers fail with vague or empty error messages, models often engage in wasteful blind retries, depleting token budgets without resolving the underlying issue. The MCP Error Relay addresses this by acting as a lightweight MCP server that provides LLMs with access to detailed server-side error logs. This crucial information allows LLMs to understand the root cause of failures, such as permission issues, and respond intelligently—whether by self-correcting, informing the user, or providing targeted solutions—instead of endlessly retrying broken calls.
Key Features
01Relays detailed error logs from other MCP servers to LLMs
02Prevents wasteful, blind LLM retries on unclear errors
03Enables intelligent and context-aware error handling by LLMs
04Lightweight and easily configurable as an MCP server
050 GitHub stars
Use Cases
01Improving LLM's ability to diagnose and communicate specific reasons for tool call failures
02Reducing token expenditure by minimizing redundant LLM retries on problematic tool calls
03Providing actionable diagnostic information (e.g., missing permissions) to LLMs for better user interaction or self-correction
