#Evidence Based Prompting MCP Servers

An AI dumped 80,000 tokens into a prompt — 64,000 of them unreferenced noise. It said 'best practice' to justify the structure and 'looks good' to measure quality. That is not context engineering. That is a copy-paste pipeline. This tool forces five context axes: relevance auditing, priority structuring, token budgeting, evidence grounding, and quality measurement.