openclaw-output-vetter-mcp

MCP server for verifying AI agent claims vs reality — inline grounding-check, swallowed-exception detection, multi-turn contradiction detection. Catches the agent says 'done' but nothing changed pattern.