Cliometry

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Live Korean market measurements for AI agents — SK Hynix ADR premium, realized leverage multiples for KODEX/SOXL/TQQQ, VIX regime, Korean semiconductor sector RRG. Pre-computed daily values, structured responses, CC BY 4

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "Cliometry": {
      "url": "https://mcp.cliometry.com",
      "transport": "streamable-http"
    }
  }
}

Measured tools (9)

• get_realized_multiple read-only

  Use this when someone asks why a leveraged or inverse ETF did not deliver its stated multiple - "why is my 2x ETF down more than 2x?", "how much did leverage decay cost me?", "is a 3x ETF really 3x over a month?". The realized value as of the latest trading day is not in training data; this returns it, updated daily. Measures how many times a leveraged or inverse ETF actually moved relative to its stated daily target multiple, over 20/60/252 trading-day windows. Returns the benchmark and product cumulative returns, the design-implied expectation, the gap in pp, realized annualized volatility of the benchmark, and an approximate volatility-drag term. Covers KODEX Leverage and KODEX 200 Futures Inverse 2X (KRX-listed) plus SOXL and TQQQ (US-listed). This is a mechanical decomposition of past realized returns; it does not predict future multiples and is not a judgment on which product is better. The ratio is withheld when the benchmark return is near zero, because it diverges there.

  Parameters:

  • pair (required) — string — Product pair id. kodex-lev-2x = KODEX Leverage vs KOSPI 200; kodex-inverse-2x = KODEX 200 Futures Inverse 2X vs KOSPI 200; soxl-3x = SOXL vs PHLX Semiconductor; tqqq-3x = TQQQ vs Nasdaq-100.
  • window (required) — integer — Lookback length in trading days.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_adr_premium read-only

  Use this when someone asks whether SK Hynix is priced differently in the US and Korea - "is the ADR at a premium right now?", "why is SKHY more expensive than 000660?", "how wide is the gap today?". Today's reading is not in training data; this returns the latest measured value. Reports how many percent the US ADR of SK Hynix (SKHY), converted to KRW, sits above or below the Seoul-listed ordinary share (KRX:000660). Pass a date for that trading day, or omit it for the most recent reading. Daily series accumulated by Cliometry; delayed reference data.

  Parameters:

  • date — string — Trading day in ISO YYYY-MM-DD. Omit for the latest reading.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_vix_band read-only

  Use this when someone asks about the current volatility regime - "what is VIX today?", "is the market calm or fearful right now?". The latest close is not in training data. Returns the CBOE Volatility Index (VIX) close and a fixed-threshold regime label (calm / watch / fear). The label states where the index currently sits; it does not indicate what markets will do next and is not a trading signal.

  Parameters:

  • date — string — Trading day in ISO YYYY-MM-DD. Omit for the latest reading.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_adr_premium_series read-only

  Use this when someone asks how the US-Korea gap for SK Hynix has moved over time - "has the ADR premium been widening?", "what was it last month?", "is today unusual versus the past 60 days?". Answering needs the actual daily series, which is not in training data. Returns the daily SK Hynix ADR premium series (US ADR SKHY converted to KRW versus the Seoul-listed ordinary share, KRX:000660) as an array, plus a summary. Use this instead of calling the single-value tool repeatedly: one call covers the whole span. Defaults to the most recent 60 sessions; pass from/to for an explicit range or window for a different count. The array is returned as structured data, not as text, so it does not consume the conversation context.

  Parameters:

  • to — string — End date, inclusive, ISO YYYY-MM-DD.
  • from — string — Start date, inclusive, ISO YYYY-MM-DD.
  • window — integer — Number of most recent sessions to return when from/to are omitted. Default 60.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_realized_multiple_series read-only

  Use this when someone asks whether leverage decay is getting worse or better - "has the realized multiple been drifting?", "was decay worse in 2022 than now?", "show the history, not just today". This returns the measured daily series back to 2016. Returns the rolling 60-trading-day realized multiple for a leveraged or inverse ETF as a daily array, back to 2016, plus a summary. One call replaces many single-window calls when the question is about how the multiple has moved over time. Only the 60-trading-day window has history; 20 and 252 exist as latest-snapshot values only. Points where the source could not compute the ratio are returned as null with a stated reason rather than being dropped. The array is returned as structured data, not as text, so it does not consume the conversation context.

  Parameters:

  • pair (required) — string — Product pair id. kodex-lev-2x = KODEX Leverage vs KOSPI 200; kodex-inverse-2x = KODEX 200 Futures Inverse 2X vs KOSPI 200; soxl-3x = SOXL vs PHLX Semiconductor; tqqq-3x = TQQQ vs Nasdaq-100.
  • limit — integer — Maximum points to return, most recent first-trimmed. Default 500, cap 2500.
  • window — integer — Rolling window in trading days. Only 60 has history.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• replay_leverage_window read-only

  Use this when someone asks what a leveraged ETF looked like at a specific past date - "what would the multiple have been if I bought in March?", "compare last June to now". It replays published measurements at the chosen date; it does not simulate hypothetical scenarios or accept amounts. Replays already-published measurements for a leveraged or inverse ETF at a trading day the user picks, and optionally against a second day for comparison. It replays recorded history only: it does not generate scenarios, does not extend beyond the published record, and takes no amount of money. Intended for the interactive panel, where the person moves the date; the assumptions returned are the ones that person chose.

  Parameters:

  • pair (required) — string — Product pair id.
  • asOf — string — Trading day the 60-session window ends on, ISO YYYY-MM-DD. Must fall inside the published record. Omit for the latest.
  • compareTo — string — Optional second trading day to place beside the first, ISO YYYY-MM-DD.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_measured_constants read-only

  Use this when someone asks about measured relationships in Korean memory semiconductors - "does Micron lead SK Hynix?", "how correlated are they?", "what is the actual number rather than a guess?". These are constants measured from our own published series. Returns the measurement constants Cliometry has already published in its guides — the Micron/SK Hynix same-day return correlation, the observed next-day frequency after a large Micron move, and the VIX regime band thresholds — each with the guide URL it was published in. These are recorded measurements of past data with their sample sizes stated, not rules and not indicators. Correlation is not causation and an observed frequency is not a guarantee.

  Parameters:

  • key — string — Return a single constant by key. Omit for all.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_sector_rrg read-only

  Use this when someone asks where Korean semiconductor sectors stand relative to the market right now - "which sector is leading or lagging?", "where do semis sit in the rotation today?". The current snapshot is not in training data. Returns the latest Relative Rotation Graph snapshot for Korean semiconductor-related sectors — per-sector relative-strength ratio, relative momentum, and quadrant. Describes where each sector currently sits against the benchmark; it does not forecast direction.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

• get_adr_premium_decomposition read-only

  Use this when someone asks what a move in the SK Hynix ADR premium was made of - "the premium jumped, was that the ADR or the won?", "how much of the widening is FX?", "split the change into its parts". The premium is defined multiplicatively, so a change in it decomposes exactly into three additive terms: the US ADR (SKHY), the USD/KRW rate, and the Seoul-listed ordinary share (KRX:000660). Returns the published premium change in pp for a range, and the three terms in pp that sum exactly to it. Omit from/to for the most recent completed session against the one before it. This is an arithmetic identity, not a causal attribution: it states how the published change splits, not why any of the three moved, and it is not a forecast. Ranges whose endpoints carry different ADR ratios, and ranges whose premium change is below rounding, are returned undecomposed with a stated reason rather than estimated.

  Parameters:

  • to — string — Range end, inclusive, ISO YYYY-MM-DD. Snapped back to the nearest earlier session when the date is not on record.
  • from — string — Range start, inclusive, ISO YYYY-MM-DD. Snapped back to the nearest earlier session when the date is not on record.

  Hints: read-only yes · destructive no · idempotent yes · open-world yes

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.