// mcp server

KALEI in Claude Code & Cursor

Ten cognitive-profiling tools, one MCP server. Ask your editor about the leaderboard, compare models, or start a BYOK profile run, without leaving the chat.

@kalei-ai/mcp-server10 toolsstdio transportMCP v1
01

Install the MCP Server

The KALEI MCP server is a Node.js package. Install globally via npm, or run directly with npx.

# Install globally
npm install -g @kalei-ai/mcp-server

# Or use npx (no install)
npx @kalei-ai/mcp-server
02

Get Your API Key

Create a KALEI API key from Dashboard → API Keys. The MCP server uses it to authenticate against kaleiai.com/api/v1.

kalei_live_xxxxxxxxxxxxxxxxxxxx
03

Configure Your MCP Client

Add KALEI to your Claude Code or Cursor MCP config. The server runs over stdio.

// ~/.claude/mcp_settings.json  (Claude Code)
// ~/.cursor/mcp.json                (Cursor)
{
  "mcpServers": {
    "kalei": {
      "command": "npx",
      "args": ["-y", "@kalei-ai/mcp-server"],
      "env": {
        "KALEI_API_KEY": "kalei_live_xxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}
04

Use the Tools

Restart your editor. All 10 KALEI tools are now callable by the model. Ask natural-language questions, the model picks the right tool.

You: "Who's #1 on the KALEI leaderboard?"
  → model calls kalei_leaderboard()

You: "Compare Claude Opus and GPT-5.4 head to head"
  → model calls kalei_compare(model_a, model_b)

You: "Profile gpt-4o for me, here's my OpenAI key"
  → model calls kalei_profile_model(...)
  → then kalei_run_progress / kalei_run_result

Available Tools

All 10 tools auto-register with your MCP client. The model chooses when to call them.

kalei_leaderboardTop models by Cognum score with cognitive types
kalei_modelFull profile for a specific model: 10-dim breakdown + volatility
kalei_compareHead-to-head comparison across all 10 dimensions
kalei_dimensionsThe 10 cognitive dimensions KALEI measures
kalei_providersSupported BYOK providers (OpenAI, Anthropic, Google, xAI, ...)
kalei_profile_modelStart a profiling run with your own provider API key
kalei_run_progressCheck completion % and current environment
kalei_run_resultRetrieve the final cognitive profile
kalei_insightsAuto-generated research findings (Sonnet Surprise, lab signatures)
kalei_ask_parliamentQuery the Parliament of models for multi-perspective answers

What is MCP?

The Model Context Protocol is an open standard for connecting AI assistants to external tools and data sources. Think of it as USB-C for LLMs: one protocol, many servers.

The KALEI MCP server exposes our leaderboard, profiling, and parliament endpoints as tools. Your AI assistant can query them mid-conversation without copy-pasting API responses.

Prefer the raw API?

Read: Profiling API Reference →