Volatility Scaling Lab Server

Overview

Volatility Scaling Lab estimates volatility, computes return and risk metrics, draws charts, and analyzes Monte Carlo results you supply. Despite the name, it does not scale a strategy to a volatility target for you, and it does not generate new simulations. For automated workflows, prefer the single-call EWMA series calculation over the stateful estimators.

Connection

MCP URL: https://quantx-api.limex.pro/mcps/volatility-scaling-lab Select Volatility Scaling Lab in the QuantX marketplace. Use its gateway connection and QuantX authorization; see Client Setup. Supplied-data calculations need no external provider key. Market-data loading depends on Yahoo availability. Cursor configuration (.cursor/mcp.json):

{
  "mcpServers": {
    "volatility-scaling-lab": {
      "url": "https://quantx-api.limex.pro/mcps/volatility-scaling-lab",
      "headers": {
        "Authorization": "Bearer ${env:QUANTX_MCP_TOKEN}"
      }
    }
  }
}

Set the referenced environment variables before starting Cursor. For Claude Code or Python, use the client-specific format in Client Setup.

Tools

Tools	Purpose and result
`load-and-prepare-data`	Load prices and cash-rate data into a prepared return table
`compute-ewma-volatility-series`	Compute a full EWMA series from `returns`, a table/reference, or a ticker
`compute-annualized-return`	Scalar return; select `method: "arithmetic"` or `"geometric"` explicitly
`compute-annualized-volatility`	Scalar annualized volatility
`compute-sharpe-ratio`	Annualized mean/volatility ratio; does not subtract a risk-free rate
`compute-maximum-drawdown`	Largest peak-to-trough decline of a cumulative wealth index (`cumulative_returns`, typically starting at `1.0`). Result is a positive fraction (`0.55` = 55%). Do not pass period returns
`compute-volatility-tracking-error`	Compare realized volatility with its target
`metrics-to-dataframe`	Turn metric records into a table
`create-metrics-table-markdown`	Return a Markdown comparison table as text
`analyze-monte-carlo-results`	Analyze supplied simulation results
`compare-strategy-to-monte-carlo`	Compare supplied strategy and simulation outputs
`plot-cumulative-returns`, `plot-weights`, `plot-controller-states`, `plot-drawdown`	Return chart payloads

Stateful estimator tools:

Prefix	Operations
`ewmavolatility-estimator`	`-create`, `-info`, `-list`, `-delete`, `-update`, `-reset`, `-get-current-volatility`
`dual-ewmaestimator`	`-create`, `-info`, `-list`, `-delete`, `-update`, `-reset`, `-get-current-volatilities`

Use types-list for constructor signatures. Domain input schemas differ; a table accepted by one tool is not automatically the input to every metric.

Examples

Calculate EWMA From Returns

Call compute-ewma-volatility-series with synthetic decimal returns:

{
  "request": {
    "returns": [0.01, -0.005, 0.002, -0.008, 0.006, 0.003],
    "halflife": 3,
    "annualize": true,
    "periods_per_year": 252
  }
}

Use a positive integer halflife and a periods-per-year value matching the observation frequency. For a table, select return_column explicitly; the default is r_risky.

Calculate Geometric Annualized Return

Call compute-annualized-return:

{
  "request": {
    "returns": [0.01, -0.005, 0.002, -0.008, 0.006, 0.003],
    "periods_per_year": 252,
    "method": "geometric"
  }
}

This tiny series demonstrates integration, not a reliable performance estimate. Use consistent units and dates across all metrics. For an excess-return Sharpe, supply excess returns yourself; the tool does not deduct RF.

Calculate Maximum Drawdown

Call compute-maximum-drawdown with a cumulative wealth index, not period returns:

{
  "request": {
    "cumulative_returns": [1.0, 1.05, 0.98]
  }
}

The result is a positive fraction (0.0667 is a 6.67% peak-to-trough decline). Passing period returns such as [0.01, -0.005, 0.002] still returns a number, but it is not a drawdown — compound the returns into a wealth index first.

Results and Limitations

Series, metric tables, and chart payloads may be returned through result_id. Retrieve the payload using result_get on this server. Charts use an image payload with artifact_type: "image", mime: "image/png", and base64 image data. Decode the reported image data after resolving the result. For file export use the artifact tools described in Shared MCP Tools. Successive -update calls on the same handle accumulate. Prefer compute-ewma-volatility-series when you already have the full series — one call, no sequential state to keep alive. The estimator tools also accept a call with no handle, but that does not give you an estimator of your own: it falls back to a shared default instance, where -update and -reset from another session land on the same object. Create your own with -create and pass its handle on every call. There are no volatility-targeting tools here. analyze-monte-carlo-results and compare-strategy-to-monte-carlo read simulation output that you pass in; generate the simulations elsewhere. Record the return convention, frequency, halflife, sample window, and any missing observations. Pin the end date for reproducible provider requests.

Overview

MCP Setup

Dataloader Server

Fama-French Replicate Server

Statistical Factor Models Server

Jump Models Server

Wavelet Mean Reversion Server

EP Ratio Screener Server