AI Agent Grounding

Your AI Agent (Claude, GPT, custom LLM)
  │
  ├─ User asks billing question or agent needs context
  │
  ├─ Agent calls RCI via MCP Server or REST API
  │   ├─ resolve_knowledge   → Full L1-L6 billing context
  │   ├─ calculate_payment   → GPCI-adjusted Medicare payment
  │   ├─ research_code       → CPT/HCPCS lookup with AI
  │   ├─ resolve_denial      → CARC/RARC interpretation
  │   ├─ search_payer        → Payer-specific rules
  │   └─ get_billing_guide   → Form, modifiers, TOB codes
  │
  ├─ Agent receives structured, sourced data
  │
  └─ Agent responds with accurate, grounded answer

{
  "mcpServers": {
    "rci-knowledge": {
      "type": "http",
      "url": "https://mcp.rci.health/v1/mcp",
      "headers": {
        "X-API-Key": "YOUR_KEY"
      }
    }
  }
}

import httpx

class RCITool:
    """Tool for LLM agents to query billing knowledge."""

    def __init__(self, api_key: str):
        self.client = httpx.AsyncClient(
            base_url="https://api.rci.health",
            headers={"X-API-Key": api_key},
        )

    async def resolve(self, ccn: str, cpt: str, payer: str = "Medicare"):
        """Resolve complete billing context for a facility + service."""
        resp = await self.client.post("/v1/knowledge/resolve", json={
            "ccn": ccn, "cpt": cpt, "payer": payer,
        })
        return resp.json()

    async def research_code(self, query: str, ccn: str = None):
        """AI-powered code research from natural language."""
        body = {"query": query}
        if ccn:
            body["ccn"] = ccn
        resp = await self.client.post("/v1/agents/code-research", json=body)
        return resp.json()  # Returns job_id, poll for result

    async def payment(self, ccn: str, cpt: str):
        """Calculate expected Medicare payment."""
        resp = await self.client.post(
            "/v1/knowledge/payment-calc",
            json={"ccn": ccn, "cpt": cpt},
        )
        return resp.json()