augure-mcp

https://github.com/log0u7/augure

Three read-only tools: analyze_target (facts in, decision + provenance out), list_rules, explain_technique. The LLM proposes, augure disposes - the decision layer every security agent can call.