Installation

Python

pip install kardashev

Requires Python 3.9+. Core dependencies: httpx, requests, pandas, numpy, pytz, openpyxl, lxml.

Optional environment variables

Only needed for direct ISONE LMP (and the legacy PJM DataMiner2 CSV path, which is dead as of 2026-07):

export ISONE_USERNAME=...
export ISONE_PASSWORD=...
 
# Legacy only - DataMiner2 CSV feeds are decommissioned
export PJM_USERNAME=...
export PJM_PASSWORD=...
 
# Optional override for api.pjm.com (otherwise the public DataMiner2 UI subscription key is used)
export PJM_API_KEY=...

ISO-NE credentials are a free account with that ISO. No credentials are required for any other dataset or ISO. For PJM LMP today, prefer the hosted Client (see Data Sources).

Verify the install

import kardashev
print(kardashev.__version__)

JavaScript / TypeScript

npm install kardashev

Requires Node.js 18+. No optional dependencies, no environment variables - the JS client only talks to the hosted API, which needs no credentials.

Verify the install

import { Client } from "kardashev";
 
const kl = new Client();
console.log(await kl.carbonLatest());

MCP (AI agents)

npx -y kardashev-mcp

Stdio Model Context Protocol  server for Cursor, Claude Code, and other MCP clients. Wraps the hosted API — LMP, carbon, load, curtailment, interconnection queues, ERCOT large loads, and the live ERCOT RT−DA spread forecast + track record. No API key.

Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "kardashev": {
      "command": "npx",
      "args": ["-y", "kardashev-mcp"]
    }
  }
}

Claude Code: claude mcp add kardashev -- npx -y kardashev-mcp.

Source: kardashev-lab/kardashev-mcp . Registry: io.github.kardashev-lab/kardashev-mcp.