A Constitutional AI governance layer that enforces 13 hard invariants (F1–F13) across agent workflows. It exposes 13 MCP tools spanning session init, evidence fetching, symbolic reasoning, and a three-tier action gate that blocks high-risk operations until they pass through an 888_JUDGE deliberation and get sealed in VAULT999. Every mutating action requires explicit human acknowledgment at Tier 2+, and irreversible ops hit a constitutional checkpoint before execution. Built for scenarios where you need traceable, reversible-by-default agent behavior with forensic audit trails. The Trinity architecture (SOUL/MIND/BODY) routes tasks across federated organs like A-FORGE for execution and GEOX for earth intelligence. You'd reach for this when agent autonomy needs to be constitutionally bounded, not just prompt-guided.
claude mcp add --transport stdio io.github.ariffazil-arifos-mcp uvx arifos-mcp