MarketHeist MCP server

Analyze portfolios that hold other portfolios and strategies, and backtest trading rules, on real price history.

MarketHeist is a free, public, remote MCP server. Put whole portfolios, a trend-filtered strategy or a rotation inside one portfolio, add a filter on top, and get CAGR, drawdown and Sharpe from real price history with caveats. It also backtests indicator rules on any ticker and decomposes funds into factors.

Endpoint: https://api.marketheist.io/api/mcp (Streamable HTTP)
Auth: none. No account, no API key.
Tools: four, all read-only.
Docs: https://marketheist.io/docs/reference/mcp
Privacy: https://marketheist.io/docs/privacy
Registry name: io.marketheist/backtest in the official MCP Registry

This repository holds the connection instructions and manifests only. The server is a hosted service, so there is nothing to install or build.

Connect

There is nothing to install. Add the URL to your client.

Claude Code

claude mcp add --transport http marketheist https://api.marketheist.io/api/mcp

Claude (claude.ai and Desktop): Settings → Connectors → Add custom connector → paste https://api.marketheist.io/api/mcp.

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

{
  "mcpServers": {
    "marketheist": { "url": "https://api.marketheist.io/api/mcp" }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "marketheist": { "type": "http", "url": "https://api.marketheist.io/api/mcp" }
  }
}

Cline (cline_mcp_settings.json)

{
  "mcpServers": {
    "marketheist": {
      "type": "streamableHttp",
      "url": "https://api.marketheist.io/api/mcp",
      "disabled": false,
      "autoApprove": []
    }
  }
}

Gemini CLI

gemini extensions install https://github.com/marketheist/mcp

Or add it to ~/.gemini/settings.json:

{
  "mcpServers": {
    "marketheist": { "httpUrl": "https://api.marketheist.io/api/mcp" }
  }
}

Clients that only speak stdio (bridge through mcp-remote)

{
  "mcpServers": {
    "marketheist": { "command": "npx", "args": ["-y", "mcp-remote", "https://api.marketheist.io/api/mcp"] }
  }
}

Anything else: Streamable HTTP at https://api.marketheist.io/api/mcp, JSON-RPC over POST, no authentication.

What makes it different

Holdings can be portfolios or strategies, not only funds. Put Golden Butterfly and All Weather inside one portfolio next to QQQ, make the stock half of a 60/40 a trend-filtered S&P 500, or hold a rotation that picks the strongest of SPY, EFA and BND each month.
Filters on top. A trend filter can wrap the whole book (hold it only while it is above its own 10-month average, else cash) or one holding.
Same numbers as a plain portfolio. Nested results come with the same CAGR, drawdown, Sharpe, risk split, caveats and reproduce hash.
The baseline is still there. Plain buy-and-hold and rebalanced mixes (16 built-in lazy portfolios or any tickers) work as usual.

Tools

Tool	What it does
`list_indicators`	Lists the 12 built-in technical indicators (RSI, Bollinger %B, MA and EMA crossover, ADX, ATR %, CCI, OBV, Stochastic, VWAP deviation, MA curvature, volume ratio) with parameters and default rules.
`run_backtest`	Backtests one indicator rule on any ticker (daily, weekly or monthly) against buy-and-hold: Sharpe, CAGR, max drawdown, Calmar, Sortino, Omega, curves and a validity block with an overfit check.
`analyze_portfolio`	Analyzes a lazy portfolio, any ticker mix, or a multi-strategy portfolio whose holdings are whole portfolios, backtested strategies or rotations, with an optional trend filter on the whole book: CAGR, drawdown, Sharpe, risk drivers.
`decompose_factors`	Regresses a ticker's returns on market, size, value, momentum, quality, low-volatility, duration and credit factors: betas, t-stats, variance shares and alpha.

Try asking

"Build me a portfolio of portfolios: 40% Golden Butterfly, 40% All Weather and 20% QQQ, rebalanced yearly. Then add a 10-month trend filter on the whole thing."
"Make the stock side of a 60/40 a trend-following S&P 500 (in SPY above its 10-month average, else cash) and keep the bonds plain. Compare with an ordinary 60/40."
"Analyze the All Weather portfolio with a 10-month trend filter. How much did it cut the drawdown?"
"I hold 60% VTI, 20% VXUS and 20% BND, rebalanced yearly. How has that done, and where does the risk come from?"
"Backtest a 50/200-day golden cross on SPY with daily data and 5 bps costs, against buy-and-hold."
"What is ARKK actually exposed to? Decompose its returns into factors."

How the numbers are made

Every number is computed by a deterministic engine on real price data (Yahoo Finance adjusted closes, dividends reinvested), not estimated by the model. Each result carries a validity block: data provenance and window, known caveats (single historical path, no costs or taxes unless set, proxy-extended history, statistical significance, a parameter-overfit check) and a reproduce hash.

Historical simulation for research and education, not investment advice.

Contact

[email protected] · https://marketheist.io

MarketHeist MCP server