Skip to content

sh.kwant/kwant

by kwant.sh · remote · no sign-in to list tools · v1.0.0 · published by sh.kwant

Quant intelligence over MCP: backtest, signals, screens, scores & portfolios for US & TSX stocks.

Changes · 12
CompareRegistry entry

16 tools: 12 change things and 4 only read.

Scanned Oct 21, 2026

No check matched anything in the tool definitions. No finding is not a safety guarantee. How the checks work.

claude mcp add --transport http kwant https://kwant.sh/mcp

Tools

Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.

16 tools

Changes things 12Claude Code: ask
  • backtestBacktest a simple long-only technical strategy on daily price history.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Backtest a simple long-only technical strategy on daily price history. Strategies: sma_cross (golden/death cross of SMA 50/200), rsi_reversion (enter RSI<30, exit RSI>70), macd_cross (MACD line vs signal). No-lookahead: signals act on the next bar's close. Returns trades + performance metrics. (paid: $0.0100/call)
  • build_monthly_universeRank a universe of tickers by monthly dollar volume with trailing returns.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Rank a universe of tickers by monthly dollar volume with trailing returns. Resolves either a named universe (SP500, TSX) or an explicit tickers override, bulk-fetches daily OHLCV over range, resamples each to monthly bars (open=first, high=max, low=min, close=last, volume=sum), and per ticker computes trailing returns over 1/3/6/12 periods plus the latest monthly dollar volume (close * volume). Tickers are ranked by latest dollar volume (descending) and the top top_n are returned. Tickers that fail to fetch or lack enough monthly history are skipped and noted in warnings. range is the lookback (5d,1mo,3mo,6mo,1y,2y,5y,max; default 2y); use a multi-year window so the 12-month horizon is well-defined. values holds range, as_of (latest monthly date, YYYY-MM-DD), count, and results — a ranked list of {ticker, dollar_volume, returns: {1m,3m,6m,12m}, rank} where any horizon longer than the available history is null. (paid: $0.0100/call)
  • compute_correlation_matrixCompute the pairwise return-correlation matrix for a list of tickers.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Compute the pairwise return-correlation matrix for a list of tickers. Fetches each ticker's daily history over range, converts it to daily returns, and computes the pairwise Pearson correlation (aligned on shared dates). Requires at least two tickers; tickers that cannot be fetched are dropped and noted in warnings (at least two must survive). Returns the standard envelope; values holds range, the tickers used, and matrix — a nested dict {rowTicker: {colTicker: correlation}} with a 1.0 diagonal. (paid: $0.0100/call)
  • compute_indicatorCompute a technical indicator (RSI, MACD, SMA, EMA, BBANDS, ATR, ADX, STOCH) over a ticker's price history.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Compute a technical indicator (RSI, MACD, SMA, EMA, BBANDS, ATR, ADX, STOCH) over a ticker's price history. Returns the warmup-aligned series plus the latest values and a one-line summary. Tune the window with `length` (SMA/EMA/RSI/ATR/ADX/BBANDS), `fast`/`slow`/`signal` (MACD), `std` (BBANDS), or `k`/`d`/`smooth_k` (STOCH) — pass them either as top-level fields OR nested under `params`; both work. `window` and `period` are accepted as aliases for `length`. NOTE: for a long window like SMA(200) you MUST set length=200 (the default is 20). The response echoes the effective params it used, and the summary shows the window, e.g. `SMA(200)`. Ensure `range` spans at least `length` bars (e.g. range=2y for SMA(200)) or the series is all-warmup NaN. (paid: $0.0050/call)
  • compute_portfolio_statsCompute portfolio-level statistics for a weighted basket of tickers.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Compute portfolio-level statistics for a weighted basket of tickers. Given a {ticker: weight} mapping, fetches each ticker's daily history over range and returns the portfolio-level (not per-ticker) volatility, sharpe, max_drawdown and total_return of the weighted basket. weights need NOT sum to 1 (normalized internally). Tickers that cannot be fetched are dropped, a note is added to warnings, and the remaining weights are renormalized. risk_free_rate is an annual rate used only by Sharpe. Returns the standard envelope; values holds range, the normalized weights used, and the stats dict. (paid: $0.0100/call)
  • compute_statsCompute quantitative statistics (volatility, sharpe, max_drawdown, returns, beta, correlation) over a ticker's daily price history.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Compute quantitative statistics (volatility, sharpe, max_drawdown, returns, beta, correlation) over a ticker's daily price history. Omit `metrics` to default to volatility/sharpe/max_drawdown/returns. `beta` and `correlation` require a `benchmark` ticker; `risk_free_rate` is used only by the Sharpe ratio. (paid: $0.0050/call)
  • compute_universe_scoresScore and rank a universe of tickers by a cross-sectional signal.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Score and rank a universe of tickers by a cross-sectional signal. Resolves either a named universe (SP500, TSX) or an explicit tickers override, bulk-fetches daily price history over range, computes a raw per-ticker score for the chosen signal, then converts those raw scores into cross-sectional z-scores and ranks them across the universe (rank 1 = highest z-score). Signals (Jegadeesh–Titman momentum is 12-month minus 1-month return on month-end resampled closes): jt_momentum (that JT 12-1 momentum), mean_reversion (negative trailing 1-month monthly return), rsi_filtered_momentum (JT momentum, names with a 14-day simple RSI > 70 excluded before z-scoring), trend_quality (JT momentum, names trading at or below their 200-day SMA excluded). Tickers with too little history or that fail to fetch are dropped and reported in warnings. range is the lookback (5d,1mo,3mo,6mo,1y,2y,5y,max; default 2y); top_n truncates the ranked output. values holds universe, signal, range, scored, and results (a list of {ticker, raw, zscore, rank} ordered by rank). (paid: $0.0100/call)
  • construct_portfolioTurn a {ticker: score} mapping into long-only portfolio weights.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Turn a {ticker: score} mapping into long-only portfolio weights. Selects names and assigns non-negative weights that sum to 1.0 using the chosen method: top_n_weighted (weight by clipped score), equal_weight, risk_parity (inverse-volatility), concentrated_vol (highest-vol from a top-score pool), or sharpe_optimized (max-Sharpe long-only). The last three fetch daily history over range (5d,1mo,3mo,6mo,1y,2y,5y,max) and convert it to returns; tickers that fail to fetch are dropped with a warning. Returns the standard envelope; values holds method, top_n, a weights map, and n_holdings. (paid: $0.0100/call)
  • detect_signalsDetect classic technical-analysis signals on a ticker's price history.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Detect classic technical-analysis signals on a ticker's price history. Each requested signal is evaluated and reported as triggered/not-triggered with a date and human-readable detail under signal_summary. Signals (omit `signals` to check all six): golden_cross = SMA(50) crosses above SMA(200) within `lookback`; death_cross = SMA(50) crosses below SMA(200); macd_cross = MACD line crosses above its signal line (bullish); rsi_oversold = RSI(14) below 30 at the latest bar; rsi_overbought = RSI(14) above 70 at the latest bar; breakout = latest close exceeds the highest high of the prior 20 bars. A signal needing more history than is available is returned not-triggered with an insufficient-history detail and a warning — it never fails the call. (paid: $0.0050/call)
  • run_portfolio_backtestBacktest a rebalanced, multi-ticker, long-only quant portfolio.No findings

    Why it is in this group

    • The name "run_portfolio_backtest" says "run": a word for a change.
    Description, as the server wrote it
    Backtest a rebalanced, multi-ticker, long-only quant portfolio. Fetches daily history for every ticker over range, then runs a walk-forward simulation: at each period-end rebalance the chosen signal (jt_momentum, mean_reversion, rsi_filtered_momentum, trend_quality) scores each name using only data up to that date, and method turns those scores into long-only weights. Returns gross and net (after cost) performance. rebalance is M (monthly) or Q (quarterly); cost_bps is round-trip cost on turnover; benchmark drives the hit-rate metric and (with crash_filter) a regime filter holding cash when the benchmark trailing-12m return is negative. Returns the standard envelope; values holds equity_curve, rebalances, metrics, holdings and the echoed parameters. (paid: $0.0100/call)
  • screenScreen a stock universe for tickers matching quantitative filters (logical AND).No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Screen a stock universe for tickers matching quantitative filters (logical AND). Fields: price, rsi, sma_50, sma_200, volatility, sharpe, max_drawdown, total_return, dollar_volume, garman_klass_vol. Ops: lt, lte, gt, gte. (paid: $0.0100/call)
  • screen_with_scoresRank a stock universe by a continuous cross-sectional signal score (rank 1 = highest z-score).No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    Description, as the server wrote it
    Rank a stock universe by a continuous cross-sectional signal score (rank 1 = highest z-score). Signals: jt_momentum, mean_reversion, rsi_filtered_momentum, trend_quality. Scores are relative to the scanned set. (paid: $0.0100/call)
Only reads 4Claude Code: allow
  • compare_tickersRank two or more tickers against each other by a single metric (total_return, volatility, sharpe, max_drawdown, last_price).No findings

    Why it is in this group

    • The name "compare_tickers" says "compare": a word for reading.
    Description, as the server wrote it
    Rank two or more tickers against each other by a single metric (total_return, volatility, sharpe, max_drawdown, last_price). Symbols that cannot be resolved (or lack enough history) are skipped and noted in `warnings` rather than failing the call. Returns a ranked list of {ticker, value, rank, currency} (rank 1 = best). (paid: $0.0050/call)
  • get_fundamentalsGet fundamental data for a ticker (profile + key ratios).No findings

    Why it is in this group

    • The name "get_fundamentals" says "get": a word for reading.
    Description, as the server wrote it
    Get fundamental data for a ticker (profile + key ratios). US symbols are bare (AAPL); TSX symbols use the Yahoo .TO form (RY.TO) or the TSX:RY form. Returns an envelope whose values holds available fundamentals: name, exchange, currency, sector, industry, market_cap, pe_ratio, forward_pe, eps, dividend_yield, beta, fifty_two_week_high, fifty_two_week_low, asof. Fields not covered by the provider are null. (paid: $0.0050/call)
  • get_price_historyGet historical OHLCV price bars for a ticker.No findings

    Why it is in this group

    • The name "get_price_history" says "get": a word for reading.
    Description, as the server wrote it
    Get historical OHLCV price bars for a ticker. US symbols are bare (AAPL, MSFT); TSX symbols use the Yahoo .TO form (RY.TO) or the TSX:RY form. interval is one of 1m,5m,15m,30m,1h,1d,1wk,1mo (default 1d); range is one of 5d,1mo,3mo,6mo,1y,2y,5y,max (default 1y). Returns an envelope whose values contains interval, range, currency, count, and bars (records with an ISO timestamp plus open, high, low, close, volume). (paid: $0.0050/call)
  • get_quoteGet the latest available quote for a ticker.No findings

    Why it is in this group

    • The name "get_quote" says "get": a word for reading.
    Description, as the server wrote it
    Get the latest available quote for a ticker. US symbols are bare (AAPL); TSX symbols use the Yahoo .TO form (RY.TO) or the TSX:RY form. Returns an envelope whose values holds the quote fields (price, currency, previous_close, change, change_percent, volume, market_state, asof as an ISO string). (paid: $0.0050/call)

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Install

Pinned to version 1.0.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http kwant https://kwant.sh/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__kwant__get_price_history",
      "mcp__kwant__get_quote",
      "mcp__kwant__get_fundamentals",
      "mcp__kwant__compare_tickers"
    ],
    "ask": [
      "mcp__kwant__compute_indicator",
      "mcp__kwant__detect_signals",
      "mcp__kwant__compute_stats",
      "mcp__kwant__screen",
      "mcp__kwant__backtest",
      "mcp__kwant__screen_with_scores",
      "mcp__kwant__compute_universe_scores",
      "mcp__kwant__build_monthly_universe",
      "mcp__kwant__construct_portfolio",
      "mcp__kwant__run_portfolio_backtest",
      "mcp__kwant__compute_portfolio_stats",
      "mcp__kwant__compute_correlation_matrix"
    ],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as kwant. Review it before you use it.