Model Context Protocol · no API key

MCP server

Give your AI assistant our backtesting engine. One streamable-HTTP endpoint turns Claude, Cursor or any MCP client into a crypto-strategy analyst with live engine data — signals, buy-and-hold history, the Overfitting Index, head-to-heads, and a real backtest runner.

https://noonbarbari.xyz/mcp

Install

Claude Code

claude mcp add --transport http noonbarbari https://noonbarbari.xyz/mcp

Claude Desktop / claude.ai (custom connector)

https://noonbarbari.xyz/mcp

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "noonbarbari": { "url": "https://noonbarbari.xyz/mcp" }
  }
}

Any Streamable-HTTP MCP client

POST https://noonbarbari.xyz/mcp
Content-Type: application/json

{"jsonrpc":"2.0","id":1,"method":"tools/list"}

Tools

list_strategies

The 10 strategy templates usable in other tools.

get_coin_signals

Today's RSI/MACD/SuperTrend/trend readings — one coin or the whole 50-coin board.

get_buy_hold

What $1,000 bought on Jan 1 of each year would be worth today, with peak and max drawdown.

get_overfitting_index

The monthly Crypto Overfitting Index, current + history.

compare_strategies

Head-to-head of any two templates from real monthly engine runs across ~50 coins.

search_answers

Search the Q&A knowledge base (backtesting, validation, risk, crypto markets).

run_backtest

Run a real BTC/USDT backtest of a template from a start date — returns return, drawdown, trades and a robustness verdict with a shareable link. Rate-limited.

Things to ask once connected

After installing, your assistant can answer these with real engine data:

"What are today's readings for BTC?"
"Compare SuperTrend vs EMA crossover on real data."
"Run a SuperTrend backtest on BTC from 2022 — is it curve-fit?"
"What if I'd bought $1,000 of SOL in 2021?"

Fair use

Free, no key, read-only. Data is CC BY 4.0 — have your application link to noonbarbari.xyz where the data is displayed. The backtest tool shares the public backtester's rate limit. No uptime guarantee; heavy automated abuse gets blocked.

All data describes past periods and is educational — not financial advice or trade signals. Past performance does not guarantee future results.