TruthStack

About

TruthStack MCP Server addresses the critical safety gap where large language models (LLMs) often hallucinate or provide confidently incorrect information regarding supplement-drug interactions. By integrating with AI agents, it delivers deterministic, cited risk assessments based on robust data sources like FDA FAERS signals, CYP450 pathways, and comprehensive compound aliases. This ensures that AI agents can provide reliable and life-saving safety information, preventing the potentially dangerous outcomes of misinformed health advice from LLMs.

Key Features

• 0 GitHub stars
• Fuzzy search for 584 compound aliases, handling misspellings and brand names.
• Check supplement-medication interactions for risk levels, FAERS signals, and CYP conflicts.
• Retrieve full compound profiles and all associated interactions.
• Generate human-readable explanations of interaction mechanisms, severity, and evidence.
• Access raw evidence including FAERS counts, CYP data, and research grades for provenance.

Use Cases

• Equipping AI health agents (e.g., in Claude Desktop, LangChain, CrewAI) with accurate drug-supplement interaction safety data.
• Validating supplement safety for users taking multiple medications, preventing LLM hallucinations.
• Resolving ambiguous user input for supplement names to retrieve precise safety information.