Aaa Mcp
Wire constitutional guardrails and tri-witness consensus into your coding agent so high-risk LLM actions stay inside enforced F1–F13 floors.
Overview
io.github.ariffazil/aaa-mcp is a MCP server for the Ship phase that enforces 13 constitutional floors (F1–F13) and tri-witness consensus on LLM agent actions.
What is this MCP server?
- 13 constitutional enforcement floors labeled F1–F13 for LLM and tool behavior
- Tri-witness consensus model for contested or high-stakes agent decisions
- stdio (PyPI arifos) and remote SSE transport for Claude Code–style hosts
- Pairs with arifOS repository for constitutional AI governance patterns
- Server schema version 60.0.0 in the official MCP registry manifest
- 13 enforced constitutional floors (F1–F13)
- Tri-witness consensus mechanism for LLM governance
- Registry version 60.0.0 with PyPI stdio and SSE remote package
What problem does it solve?
Solo builders cannot trust open-ended coding agents with sensitive repos when policy exists only in system prompts that models ignore under pressure.
Who is it for?
Indie developers deploying agentic workflows who need named constitutional floors and witness-style agreement without building a custom policy engine.
Skip if: Teams that only want lightweight linting or a single static system prompt with no MCP-hosted enforcement layer.
What do I get? / Deliverables
After registration, your agent can route governance checks through MCP so floor violations and consensus gaps surface before irreversible tool calls complete.
- Registered MCP connection with stdio or SSE transport
- Runtime access to constitutional floor and tri-witness governance tools
- Inspectable governance behavior aligned with arifOS F1–F13 model
Recommended MCP Servers
Journey fit
Governance MCPs belong on the Ship shelf because solo builders adopt them when hardening agent behavior before or alongside production release, not when first sketching product ideas. Security is the canonical subphase for enforced policy floors and witness consensus that block or gate unsafe model outputs and tool use.
How it compares
MCP constitutional governance server, not a general-purpose linter or single Claude skill markdown file.
Common Questions / FAQ
Who is io.github.ariffazil/aaa-mcp for?
It is for solo and small-team builders who run MCP-capable agents and want arifOS-style constitutional floors plus tri-witness consensus on risky steps.
When should I use io.github.ariffazil/aaa-mcp?
Use it in the Ship and security-hardening window when you are wiring guardrails around agent tools, deployments, or data access—not for initial market research.
How do I add io.github.ariffazil/aaa-mcp to my agent?
Add the server to your MCP client config using stdio with the arifos PyPI package or point remote transport at the published SSE URL from the registry manifest.