Quantamental MCP Server

MCP server exposing quantamental analysis tools — technical indicators, fundamental factors, and trading strategies — over the Model Context Protocol. Built on @quantamental/tools and the native Rust core (@quantamental/core); all tools (indicators, factors, strategies) are available to any MCP client.

Installation

bun add -g @quantamental/mcp

Or run directly without installing:

bunx @quantamental/mcp

Setup

After a global install, the quantamental-mcp binary is on your PATH. Configure it in your AI client (e.g., Claude Desktop, VS Code settings, opencode.json):

{
	"mcpServers": {
		"quantamental": {
			"command": "quantamental-mcp"
		}
	}
}

Or, without a global install, run it on demand:

{
	"mcpServers": {
		"quantamental": {
			"command": "bunx",
			"args": ["@quantamental/mcp"]
		}
	}
}

No environment variables are required — all computation runs locally against the native Rust core.

Tools

Tool names are prefixed by category (strategy_, indicator_, factor_) because 35 ids are shared by the indicator and strategy categories (e.g. rsi is both an indicator and a strategy).

| Category | Count | Input | Dispatch | | ---------- | ----- | ---------------------------------- | ----------------------------------------- | | Indicators | 177 | values: number[] | core[id] or core[camelCase(id)] | | Factors | 72 | fundamentals: FundamentalPoint[] | core[id] or core[camelCase(id)] | | Strategies | 112 | closes: number[] + OHLCV + config | core.runStrategy(id, input, config) |

Strategies

Take OHLCV arrays (closes required; highs, lows, volumes optional) and an optional config object whose keys match the strategy's params schema (camelCase, e.g. { fastPeriod: 10, slowPeriod: 20 }). Return a signal array (1 = buy, -1 = sell, 0 = hold).

Indicators

Take a values array (oldest first) and dispatch to the matching native export by id (e.g. rsi → core.rsi, awesome_oscillator → core.awesomeOscillator). Return indicator-specific result objects.

Factors

Take an array of FundamentalPoint objects and return FactorPoint[] ({ symbol, date, value }). Each fundamental point has symbol, date, filingDate, period, and a data object with financial fields (all camelCase, e.g. marketCap, shareholdersEquity, eps, revenue).

Most factors follow that shape, but a few expect different inputs, passed through the same fundamentals argument:

• 5 factors also accept a prices array of Bar objects ({ time, open, high, low, close, volume }): analyst_target_upside, dividend_yield, margin_of_safety, price_to_affo, price_to_earnings.
• 4 on-chain factors take an array of OnChainDataPoint objects ({ time, metric, value }) as fundamentals: active_address_growth, exchange_flow_momentum, nvt_ratio, staking_ratio.
• 2 prediction-market factors take an array of PredictionMarketPoint objects ({ time, marketId, price, volume }) as fundamentals: odds_momentum, prediction_market_odds.

Example:

{
	"name": "factor_price_to_book",
	"arguments": {
		"fundamentals": [{
			"symbol": "AAPL",
			"date": 1000,
			"filingDate": 1000,
			"period": "Q1",
			"data": { "marketCap": 300000, "shareholdersEquity": 150000 }
		}]
	}
}

Development

bun run typecheck
bun run test