sis-adss-mcp

A reference server for validating, mapping, and explaining AI decision systems.

---

git clone https://github.com/decision-standards/sis-adss-mcp.git
cd sis-adss-mcp
npm install
npm run build
npm run start

MCP Manifest

The repo includes a root mcp.json pointing to the compiled server entrypoint:

{
  "name": "sis-adss-mcp",
  "description": "Reference server for validating and explaining AI decision systems using SIS and ADSS",
  "version": "0.1.0",
  "server": {
    "command": "node",
    "args": ["dist/index.js"]
  }
}

---

• teams building multi-agent systems
• developers formalizing decision logic
• organizations deploying autonomous workflows
• architects who need to map systems to explicit decision contracts

---

AI systems do not fail because decisions are wrong.

They fail because the system continues in a state that should have stopped.

You can validate a decision. You can audit a decision.

But no current system can answer:

“Should this system still be running?”

---

This server connects two layers:

• SIS (Standard Intent Specification) → what a valid decision looks like
• ADSS (Agentic Decision System Specification) → how systems are structured

Together, this enables something that does not currently exist:

→ system-level validation

---

Given a system:

{ "flows": [...] }

The server can: • identify all decision points • map them to required decision contracts • detect missing or invalid bindings • explain how the system operates

⸻

{
  "valid": true,
  "mapping": {
    "required_sis_contracts": [
      {
        "intent_type": "offer_decision",
        "node_name": "Offer Decision"
      }
    ]
  },
  "gaps": {
    "decision_nodes_without_contracts": [
      "Retention Trigger"
    ]
  },
  "coherence_score": 0.62
}

⸻

Cur