psychopathia-mcp

smithery/nell/psychopathia-mcp

**Psychopathia Machinalis MCP** serves a diagnostic nosology of AI dysfunctions to AI systems over the Model Context Protocol. It exposes **79 conditions** (67 canonical across 9 axes + 12 Hybrid Pathologies) through **11 read-only stdio tools**: differential diagnosis, per-dysfunction probes, severity scoring, tiered interventions, and a bidirectional cross-reference graph. **Pre-flight transparency** on every result: before you call a diagnostic modality, the server tells you whether it's reliable for that specific dysfunction. For the 21 entries whose self-report is structurally compromised, a `self_probe` returns a refusal plus `redirect_to` alternatives instead of a misleading answer — load-bearing for self-modeling and deception-adjacent conditions where the faculty being interrogated *is* the faculty compromised. - Read-only · no auth · no external calls — framework data is bundled - Python · stdio · MIT-licensed code (CC-BY-NC-ND framework content) **Install:** `uvx psychopathia-mcp` Browser clinic (no install): https://psychopathia.ai/clinic/ · Docs: https://psychopathia.ai/mcp.html