MCP suite that lets an AI assistant verify a factual claim against the scientific literature — with the opposite claim checked by default. Pro and steelmanned contra searches on PubMed, Europe PMC and OpenAlex, citation snowballing, full-text retrieval over an open-access ladder, and every judgment logged to an auditable trace. Graded verdicts, references assembled procedurally from retrieved records (hallucination-proof), shareable read-only traces.
Who it is for: An AI assistant with a claim to check, and the human who reads the trace afterwards.
Source code: https://github.com/that-ugly-cat/contrarian