"Temporal" matching MCP servers:

Audits local datasets and ML training code for target, temporal, and entity leakage, PII exposure, and invalid evaluation design, returning structured evidence and pass/review/block verdicts before model training.

Enables users to expose temporal architectural memory, privacy sanitization, and local test-runner capabilities as MCP tools so clients like Claude Desktop, Cursor, and Hermes Agent can query decisions, sanitize prompts, and verify code changes.