Starkov MCP Protocol & Non-Custodial Live Relay (starkov-mcp v1.2.0)

Official open-source Model Context Protocol (MCP) integration and Non-Custodial Live Execution Relay (starkov-relay.mjs) for Starkov — the quantitative sports prediction market platform for Kalshi and Polymarket.

• Starkov Web Platform & Bot Studio: https://starkov-app.vercel.app/bots

• Personal MCP Key & Prompt Book: https://starkov-app.vercel.app/doc?section=mcp-copilot

• Remote Streamable HTTP MCP Endpoint: https://starkov-app.vercel.app/api/mcp

1. Quick Setup (Connect Your AI Assistant in 30 Seconds)

First, sign in to starkov-app.vercel.app/doc?section=mcp-copilot and copy your personal MCP token (stk_mcp_...).

Claude Code (CLI)

claude mcp add --transport http starkov "https://starkov-app.vercel.app/api/mcp" --header "Authorization: Bearer stk_mcp_your_personal_token_here"

Antigravity (CLI)

agy mcp add starkov "https://starkov-app.vercel.app/api/mcp" --header "Authorization: Bearer stk_mcp_your_personal_token_here"

Codex (CLI)

codex mcp add starkov --url "https://starkov-app.vercel.app/api/mcp"

Cursor / Windsurf / JSON Config (mcp.json)

{
  "mcpServers": {
    "starkov": {
      "url": "https://starkov-app.vercel.app/api/mcp",
      "headers": {
        "Authorization": "Bearer stk_mcp_your_personal_token_here"
      }
    }
  }
}

Claude Desktop (Stdio Bridge via bin/starkov-mcp-bridge.mjs)

If your MCP client only supports local stdio servers:

{
  "mcpServers": {
    "starkov": {
      "command": "node",
      "args": ["/absolute/path/to/starkov-mcp/bin/starkov-mcp-bridge.mjs"],
      "env": {
        "STARKOV_MCP_TOKEN": "stk_mcp_your_personal_token_here"
      }
    }
  }
}

Related MCP server: SportsQuant MCP Server

2. How the AI Assistant Guides You (From Idea to Live Deployment)

When connected to starkov-prediction-mcp, your AI assistant follows a structured 3-Stage Quantitative Workflow:

Step A — Interactive 4-Pillar Bot Discovery (Ask First, Never Guess)

When you ask your AI to create or optimize a bot (e.g. "Build me a tennis bot"), the AI will:

1. Call starkov_list_bots to check your current subscription tier (Starter Free Level 1, Pro $19/mo Level 2, or Quant $49/mo Level 3) and existing bots.

2. Ask you 2 to 4 concise multiple-choice questions covering any missing pillars using options unlocked on your plan:

   • 01 TARGET: Sport (Tennis, Soccer, NBA, NFL, NHL, MLB), circuits/leagues, and court surface / environment.

   • 02 TRIGGER: Entry condition (Pre-Match Value vs. Live In-Play Comeback Dip, Favorite vs. Underdog) and minimum de-vigged +EV Edge % (> 1.5%, > 2.5%, > 4.0%, > 6.0%).

   • 03 ORDER: Stake sizing ($10 / $25 / $50 Fixed USD, % of Wallet, or [Pro] Fractional Kelly Criterion), contract side (YES / NO), and order type (Limit Maker 0% Fee vs. Market Taker Instant).

   • 04 EXIT: Exit discipline (Wait Until End of Match vs. Odds Convergence Early Cashout), Take-Profit %, and Stop-Loss %.

Step B — Instant Live Dry-Run & 24/7 Cloud Paper Sandbox

Once your 4 pillars are configured via starkov_configure_bot:

• Live Market Dry-Run (starkov_evaluate_bot_on_live_markets): The AI tests your bot right now against live and upcoming matches using Python's evaluator.py + execution_simulator.py (computing L2 VWAP orderbook slippage and official Kalshi/Polymarket fees).

• 24/7 Cloud Paper Sandbox: When your bot status is active, Starkov's cloud engine monitors live games 24/7 and executes paper trades in your $1,000 sandbox wallet at https://starkov-app.vercel.app/wallet.

Step C — Deploy to Live Kalshi & Polymarket (starkov-relay.mjs)

At the end of the workflow, when you are ready to route signals from your bot to your real-money Kalshi or Polymarket account:

• Ask your AI assistant: "Prepare live deployment for my bot on Kalshi" (starkov_prepare_live_deployment), OR click the Deploy button in the top bar of https://starkov-app.vercel.app/bots.

• Choose between:

  1. 1-Click Railway Cloud Deploy: Opens the pre-filled Railway template for this repository (Dockerfile + starkov-relay.mjs). Paste your Kalshi/Polymarket keys into your own private Railway service and click Deploy.

  2. Local Terminal / VPS / Docker CLI: Run node starkov-relay.mjs directly on your own machine or server (requires Node.js 18+, zero external npm dependencies).

3. Available MCP Tools (10 Tools) & Built-in Documentation Resources

MCP Tools

Built-in MCP Resources (resources/list & resources/read)

Your AI assistant can read any of these built-in reference documents directly over MCP:

• starkov://docs/bot-builder-discovery-guide — Interactive 4-Pillar Bot Builder & User Questioning Guide

• starkov://docs/end-to-end-deploy-and-live-relay — End-to-End Bot Lifecycle & Non-Custodial Live Deployment (starkov-relay.mjs)

• starkov://docs/4-pillar-architecture-and-tiers — 64-Slot Strategy Architecture & Subscription Tiers

• starkov://docs/microstructure-vwap-and-fees — Orderbook Microstructure, L2 VWAP Slippage & Exchange Fee Math

4. Running the Non-Custodial Live Relay (starkov-relay.mjs)

Why Non-Custodial?

Starkov never asks for, sees, or stores your Kalshi RSA Private Key or Polymarket CLOB L2 keys. starkov-relay.mjs is a 100% open-source, zero-dependency Node.js 18+ script that polls your authenticated bot signal queue (STARKOV_BOT_TOKEN=stk_live_...) and signs orders locally using Node's built-in node:crypto.

Option A: 1-Click Railway Deploy

Run starkov_prepare_live_deployment in your MCP client (or click Deploy on /bots) to open the pre-filled Railway template, paste your exchange credentials, and deploy in 1 click.

Option B: Local CLI / VPS / Docker

git clone https://github.com/FUTFURY/starkov-mcp.git
cd starkov-mcp
cp .env.example .env
# Edit .env with your STARKOV_BOT_TOKEN (stk_live_...) and Kalshi/Polymarket API keys
node starkov-relay.mjs

Or run with Docker:

docker build -t starkov-relay .
docker run -d --name starkov-relay --env-file .env -p 8080:8080 starkov-relay