AIxEnergy Model Context Protocol (MCP)

1. Executive Architecture & System Motivation

Developing large-scale artificial intelligence compute campuses (100 MW to 1,000+ MW) and clean generation assets requires reconciling asynchronous, geographically dispersed data across transmission physics, substation headroom, ISO study queues, environmental gates, and capital stacks.

Historically, evaluating a candidate power campus required 4 to 8 weeks of fragmented consulting across separate GIS, electrical engineering, and legal teams. The AIxEnergy Universal MCP Server collapses this diligence cycle into sub-100 millisecond edge queries, enabling AI agents and production cost models to directly interrogate the canonical AIxEnergy global project atlas and grid models.

2. Client Quickstart Configurations

The server supports the standard Model Context Protocol Streamable HTTP transport and Server-Sent Events (SSE).

Cursor Configuration

Add to your .cursor/mcp.json or configure in Cursor Settings → Features → MCP:

{
  "mcpServers": {
    "aixenergy-unified": {
      "url": "https://map.aixenergy.io/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_AIX_MCP_BEARER_TOKEN>"
      }
    }
  }
}

Claude Desktop Configuration

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "aixenergy-unified": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@latest",
        "https://map.aixenergy.io/mcp",
        "--header",
        "Authorization: Bearer <YOUR_AIX_MCP_BEARER_TOKEN>"
      ]
    }
  }
}

Python SDK & Power Modeling API Connector

import requests

url = "https://map.aixenergy.io/mcp"
headers = {
    "Authorization": "Bearer <YOUR_AIX_MCP_BEARER_TOKEN>",
    "Content-Type": "application/json"
}

# Resolve PNode and electrical bus for power systems modeling
payload = {
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
        "name": "resolve_pnode_and_electrical_bus",
        "arguments": {
            "query": "Austin, TX",
            "market": "ERCOT"
        }
    },
    "id": 1
}

res = requests.post(url, headers=headers, json=payload)
data = res.json()
print("Bus Injection Tag:", data["result"]["content"][0]["text"])

3. Authentication & RFC 9728 Compliance

AIxEnergy MCP implements RFC 9728 OAuth Protected Resource Metadata discovery at: GET https://map.aixenergy.io/.well-known/oauth-protected-resource

Supported authentication methods:

• AIxEnergy Member Bearer Tokens: Signed HMAC tokens representing active Map or Intelligence subscriptions.
• Ghost Member JWTs: Direct publication membership tokens.
• Scoped API Keys: Permanent service tokens (aix_mcp_... or aix_live_...).
• Ephemeral Demo Sandbox Tokens: Instant sandbox credentials issued via POST https://map.aixenergy.io/api/mcp/token.

4. Power Systems & Production Cost Modeling

AIxEnergy MCP v3.4.0 is optimized to serve as an authoritative, vendor-neutral grid-injection data bridge for enterprise power system simulation and production cost modeling engines:

• Production Cost Simulation Engines: Direct nodal injection bus tags, heat rates, and 8,760 chronological load shapes.
• Capacity Expansion Models: Zonal and nodal load forecasts, expansion candidates, and queue backlogs.
• Nodal Power Flow & Congestion Engines: Transmission congestion modeling, LMP basis spread shifts, and contingency testing.

5. Tool 12: resolve_pnode_and_electrical_bus

6. Tool 13: generate_parametric_8760_profile

7. Interconnection Queue Harmonization & Provenance Segregation

AIxEnergy harmonizes 3,756 interconnection queue dockets while strictly segregating genuine regulatory filings from synthetic campus placeholders:

• AUTHENTIC_RTO_DOCKET (2,837 records): Genuine filings in PJM, ERCOT, NYISO, MISO, CAISO, SPP, and public utility tariffs.
• SYNTHETIC_CAMPUS_MIRROR (919 records): AIxEnergy research campus representations tagged to prevent double-counting in grid models.

8. Flagship Tri-Domain Composite Workflows

9. Modular Domain Diligence Tools

10. Decision & Analytical Prompts Catalog (19 Prompts)

Pre-packaged institutional prompts ready for one-click invocation in Claude and Cursor:

• model_nodal_load_injection: Resolves PNode, bus voltage, and nodal export tags for power system modelers.
• generate_8760_hourly_vector: Guides parameter selection and generates thermodynamic 8760 load shapes.
• underwrite_campus_power: Full 360-degree underwriting synthesis across physical, queue, and capital data.
• screen_market_capacity: Siting and capacity screening across a specified target market or county.
• audit_queue_viability: Queue viability analysis auditing developer claims vs actual RTO dockets.
• simulate_counterfactual_delay: Grid-to-capital shock simulation evaluating COD slippage and DCF IRR impact.

11. Data Governance & "Blank is Unknown" Standard

AIxEnergy enforces strict institutional data integrity standards across all MCP endpoints:

• Blank is Unknown, Never Zero: Missing numeric fields remain null. Never coerced to 0.
• Evidence Maturity != Success Probability: Data quality grade measures documentation fitness, not economic viability.
• Epistemic Transparency: Modeled curves carry explicit MODELED_PARAMETRIC_SHAPE disclosures to distinguish them from metered utility SCADA data.
• Approximate Coordinates: Coordinates are approximate centroids for regional screening, not survey-verified parcel boundaries.