Fincept

The account's credit balance: credits (the total), granted_credits (this period's allowance, which can expire), purchased_credits (bought, never expire) and tier. Check it before starting work that costs credits.

The account's server agent runs under "runs", newest first: id, kind (task, team, workflow, desk, chat), goal, status, question (when waiting for the user), error and timestamps. Use agents_get_run for a run's plan, steps and result.

The account's algo trading strategies: id, name, timeframe, direction, entry/exit condition trees, stop loss, take profit, trailing stop (percent), position size percent, scale-in settings, eval_on and version.

Lists every Fincept Allocation tool by module: portfolio optimisation, risk parity, hierarchical clustering portfolios, estimation (expected returns, covariance, views, factor models) and portfolio risk analytics.

Hierarchical clustering portfolio: HRP, HERC, HERC2 or NCO on a tree of the assets built from a codependence (correlation, Gerber, distance correlation, mutual information, tail dependence) and a linkage (single ... ward, DBHT), allocating by one of 38 risk measures, with optional weight bounds. Returns the long-only weights, the profile, the clusters, the tree as merge steps and leaf order (dendrogram numbers) and the codependence matrix in leaf order (heatmap). Needs no expected returns, so it is stable out of sample.

The risk report of a portfolio (or one series): mean return and CAGR; standard deviation, MAD, semi deviation, first and second lower partial moments, VaR, CVaR, EVaR, tail Gini, RLVaR and worst realization, annualised, each with its (return - MAR) / risk ratio; skewness and excess kurtosis; and the ulcer index, average drawdown, DaR, CDaR, EDaR, RLDaR and maximum drawdown of uncompounded cumulative returns. The same table as a portfolio report sheet.

Geopolitical news events extracted from crawled articles, newest first: url, source, event_category, title, city, country, latitude, longitude, extracted_date, confidence, event_summary, with pagination. Costs 1.5 credits (action geo_events_query) once per distinct filter set per UTC day; other pages and repeats that day are free.

One country's sovereign row (yields in percent, spreads and CDS in bp, policy rate, ratings, curve slopes), its yield forecasts and its rank among all covered countries on 10Y yield, CDS and policy rate. Free.

Daily technical-indicator grid for a symbol (SMA, EMA, MACD, ADX, Aroon, RSI, stochastic, Williams %R, ROC, MFI, ATR, Bollinger, OBV, VWAP, CMF and more), one row per bar plus a summary with buy/neutral/sell tallies per indicator group and an overall signal. At least 120 bars are fetched whatever the range. Free; a data-level refusal comes back as success=false with an error.

QuantStats-style report over one year of daily history for a saved portfolio: performance (total and annualised return, best/worst day), risk (annualised volatility, max drawdown, daily 95% VaR and CVaR, downside deviation), ratios (Sharpe, Sortino, Calmar, profit factor) and the return distribution. Costs 2 credits (action portfolio_ai_analysis), not charged when the result carries an error (e.g. no holdings).

The engine's strategy template catalog: each template's id, name, category, description, whether it is a portfolio strategy, the data it needs and its params (name, type, default, range or options). Start here to build a spec with strategy.kind "template".

Runs one backtest of a complete spec on the native engine and returns the engine's result document with its run_id, warnings and cached flag; a document too large to inline comes back as type result_ref with the run_id to read through backtest_get_run and backtest_get_run_records. Costs 1 credit (action backtest_run), charged only when a result is delivered; an identical repeat is served from history for free. Validate first with backtest_validate_spec.

Checks a complete backtest spec exactly as a run would prepare it, free. Returns valid, issues and the engine's errors (each with a path), warnings (clamped window, trimmed universe), resolved_spec with every default the engine applied, and the estimated evaluation count for a sweep. A refused spec is an answer (valid false), not an error. Run it with backtest_run.

Lists every Fincept Change Points tool by module: search (general change-point detection, change score curve), selection (penalty sweep, number of breaks by BIC/AIC/elbow), regimes (return and volatility regimes, trend breaks, regression/beta breaks, autoregressive breaks), metrics (compare segmentations, score given break dates) and synthetic (teaching signals with known breaks). Give module to list one toolset.

Lists every Fincept Copula tool by module: pair (bivariate copula fit, evaluation, simulation, conversions, families), vine (vine copula fit, structures, evaluation, simulation), dependence (dependence and tail-dependence matrices, pseudo-observations, uniform samples), distribution (margins and joint distributions), risk (copula VaR/ES, stress scenarios) and regression (vine quantile regression). Give module to list one toolset.

Per-venue tickers for up to 50 symbols (last, bid/ask with sizes, mid, 24h open/high/low, volume, change percent, mark and index) and Fincept's cross-venue index price. Prices are strings in the quote currency. Unified symbol: BTC/USDT (spot), BTC/USDT:USDT (linear perp), BTC/USD:BTC (inverse perp), BTC/USDT:USDT-260327 (dated future).

Income statement, balance sheet and cash flow of an SEC filer as filed in XBRL, by ticker or CIK: periods (newest first), each line keyed by period, in the filing currency, plus derived lines named in derived and as_of = latest filing date. Payments are negative cash flows. Non-SEC listings (.NS, .L, .T) answer not_covered. Free.

Values of up to 50 series between two dates: per series title, unit, frequency and [{date, value}] oldest first, with status ok, fetching (a first fill was queued: call again shortly) or missing. History is clamped to the plan. Re-loading a series already loaded this month is free; NEW series count against data_series_loads_monthly and past it each call costs 0.1 credit (action data_series_load). as_of needs the point-in-time capability.

Search 1.7M+ economic and financial time series across every provider, best match first: provider_id, native_id, title, units, frequency, geo, start_date, end_date, status. A provider or country named in q becomes a filter and is reported. Use the provider_id:native_id of a row with data_get_observations. Free.

Search the SEC company master by name or ticker, or browse a sector/industry: cik, display_name, tickers, exchange, sector, industry, is_fund, is_bdc, is_active, latest_filing. Use it to turn a name into the CIK the filings tools take. Give at least one of q, sector, industry or funds=only. Free.

Run any Fincept tool by name with its arguments, including tools not listed directly. Find names with fincept_search_tools and schemas with fincept_describe_tool.

Full description and JSON input schema of one Fincept tool, by name.

Reads a result that was too large to return inline: any field by path, any list by page. The id comes from the result's artifacts[].id and truncated.lists names the cut lists (path like data.candles or blocks.forecast.rows). Free.

Find Fincept tools by what you need (e.g. 'option chain', 'insider trades', 'arima forecast', 'cointegration test'). Returns each match's name, signature, effect and what it does. Then call the tool by its name, listed or not (or through fincept_call_tool).

Searches, for each chosen model, its hyperparameters together with the features (candidate lags, EWM and rolling means, calendar features and target transforms such as log, differencing and scaling, shaped by the season) to minimize the backtest error over n_windows windows, then retrains the best configuration on all data and forecasts h periods with conformal bands. Returns each model's best loss, hyperparameters and features, the loss of every trial, the forecasts and charts. The search stops at trials or about 200 seconds.

List every Fincept Forecast tool by module: machine-learning forecasting of one series or a panel with gradient boosting and regression models, conformal prediction intervals, recursive and direct multi-step strategies, model update and transfer, backtesting, LightGBM boosting-round search, automatic hyperparameter search, feature engineering (lags, rolling statistics, calendar, Fourier, exogenous lags), panel preparation and forecast accuracy evaluation.

OHLCV candles for a symbol, oldest first. Intervals: 1m 2m 5m 15m 30m 60m 90m 1h 1d 5d 1wk 1mo 3mo. History depth is clamped to the plan's history_years, reported in the answer.

Upcoming macro releases and central-bank events in UTC, soonest first, deduplicated. Defaults to the next 14 days and medium or high impact.

Key fundamentals for an equity: valuation ratios, margins, growth, balance-sheet figures, dividend and share data as the provider reports them.

Latest quote for one symbol: price, change, change percent, day range, volume, currency, exchange time and the feed's delay. Symbols use exchange suffixes (AAPL, RELIANCE.NS, 7203.T, BTC-USD, ^GSPC, EURUSD=X).

Latest quotes for up to 300 symbols in one call. Symbols that could not be quoted are listed with the reason. Prefer this to repeated market_get_quote calls.

Search the instrument universe by ticker, name or ISIN. Returns matches with symbol, exchange, asset type and currency, best match first. Use it to turn a company name into the symbol the other market tools take.

Lists every Fincept Multi-Period tool by module: back-tests (optimisation and rule policies), trades (today's optimal or rule trades, pre-trade evaluation), tuning (aversion sweeps and searches) and forecasts (return, risk, factor and volume forecasts, data preparation). Give module to list one toolset.

Latest articles from the curated feeds, newest first: title, summary, source, url, category, published_at (UTC), sentiment label and score, tickers. Free. Filings are excluded unless category=corporate or a source is named; history is clamped to the plan's window.

The enabled news feeds: id, name, url, category, region, publisher, tier and last poll status, plus the plan's news_sources limit. The ids are what the source filter of news_get_feed and news_search_articles takes. Free.

Search the article corpus, newest first, same row shape as news_get_feed. Words in q are ANDed; commas, semicolons, | or OR separate alternatives; a quoted q is one exact phrase; ticker symbols such as RELIANCE.NS match the company. Free; takes the same filters as the feed.

Lists every Fincept Panel tool by module: panel (fixed and random effects, between, first difference, pooled, Fama-MacBeth, comparison with Hausman, absorbed high-dimensional effects), iv (2SLS, LIML, GMM and their diagnostics, comparison), system (SUR, 3SLS, system GMM), asset_pricing (factor models) and data (example datasets, panel structure and transforms, HAC bandwidth). Give module to list one toolset.

The Fincept Paper account totalled across its main book and every sleeve in the base currency: equity, cash, unrealised, today's realised, day and total P&L, margin used, open orders, positions, exposure by asset class and currency, top 5 positions, and each book's id, name, sleeve_key and equity. portfolio_id is the MAIN book, the book_id the other paper tools take. 404 account_not_found until paper_ensure_account has run.

A book's orders for one trading day (IST, default today; orders resting from an earlier session included): id, symbol, side, type, quantity, prices, filled quantity, average price, status, client_order_id. status=pending lists every open order whatever its day.

A book's open positions: id, symbol, canonical instrument, exchange, side (long/short), quantity, entry and current price (quote currency), unrealised and realised P&L and held margin (base currency), product, attached stop_loss/take_profit.

Lists every Fincept Portfolio Lab tool by module: optimisation, hierarchical portfolios, moment estimators, priors and views, validation (walk-forward, combinatorial purged CV, online learning, hyper-parameter search), portfolio analytics, distributions and stress tests, and the cross-sectional factor lab. Give module to list one toolset.

Lists every Fincept Portfolio Optimizer tool by module: estimates (expected returns, covariance and shrinkage), mean-variance (optimise, frontier, custom objectives), downside risk (semivariance, CVaR, CDaR), critical line, hierarchical risk parity, Black-Litterman, discrete allocation and portfolio analytics. Give module to list one toolset.

Hierarchical risk parity weights from a price (or returns) history, or from a covariance matrix alone: the assets are clustered by correlation distance, ordered along the tree and risk is split between sub-clusters by inverse variance, needing no expected returns and no matrix inversion. Returns weights (table + bar chart), annual return (history only), volatility and Sharpe, the linkage table, the leaf order with dendrogram coordinates (series + chart) and the correlation matrix in cluster order (heatmap).

Mean-variance optimal weights from a price history (or given expected returns and covariance): max Sharpe, min volatility, max quadratic utility, max return for a target volatility, or min volatility for a target return, with bounds, shorting, market neutrality, sector/group and linear constraints and L2, transaction-cost and tracking-error penalties. Returns cleaned and raw weights (table + bar chart) and the expected annual return, volatility and Sharpe ratio. For downside risk use pfopt_semivariance, pfopt_cvar or pfopt_cdar.

Time-weighted and money-weighted performance: period returns (1d to since inception), cumulative series, drawdown, monthly and calendar-year tables, a rolling series, full and 1-year statistics (annual return, volatility, Sharpe, max drawdown, beta, tracking error) and, with start/end, stats_custom over that window. Returns are fractions.

Everything a first look needs in one call: the header, the priced valuation (NAV, cash, P&L, exposures, positions, in the base currency), the daily NAV series, the performance headline (1d, mtd, ytd, 1y and since-inception returns, 1-year stats, money-weighted return), limit breach counts and the cached risk headline (volatility, 95% VaR/CVaR, beta). risk is null with risk_pending true while the risk report is being computed; ask again shortly.

The user's tracked portfolios (books): id, name, base currency, cost method, cash tracking, benchmark, source, version and archived flag. Start here to get the portfolio_id every other portfolio tool takes.

Lists every Fincept Pricing module and tool: dates and calendars, interest-rate maths, yield curves, inflation, credit, bonds, swaps, rate options, equity and FX options, volatility, models and calibration, and numerical tools. Pass module to list one.

Prices a European, American or Bermudan call or put on a stock, index, FX rate or commodity under Black-Scholes-Merton (flat volatility or a strike x expiry surface; continuous yield and discrete dividends; FX with rate = domestic and dividend_yield = foreign rate) with any engine: analytic, dividend models, integral, American approximations, binomial trees, finite differences, Monte Carlo. Returns the price, every Greek the engine provides (delta, gamma, vega, theta, rho, dividend rho, elasticity, strike sensitivity, in-the-money probability) and, from a market price, the implied volatility.

Lists every Fincept Quant module and tool: series transforms and algebra, calendar operations on dated series, rolling statistics and regression, returns/volatility/correlation/drawdown/Sharpe, technical indicators, basket backtests, event studies, epidemic models and business-day date tools. Pass module to list one.

Returns of a dated price series over a number of observations or a tenor (simple, logarithmic or absolute), the price path implied by a return series, the series rebased to a starting level or net of its first value, annualized by its date spacing, or its excess-return index over a cash rate or cash index. Returns the result as a dated series with a line chart. Use quant_volatility for volatility and quant_sharpe_ratio for risk-adjusted performance.

Rolling (or, with no window, expanding) statistic of one dated series over a window of observations or a tenor such as '1m' or '1y': min, max, range, mean, root mean square, median, mode, sum, product, standard deviation, variance, z-score, winsorized values, percentile, percentile rank, or exponentially weighted standard deviation. Returns the statistic as a dated series with a line chart (stat percentile with no window returns one number). For volatility of returns use quant_volatility.

Technical indicator of a dated price series: simple, Wilder-smoothed or exponential moving average, Bollinger bands (price, middle, lower, upper), relative strength index or MACD, over a window of observations or a tenor where the indicator has one. Returns the indicator as a dated series with a line chart (a band chart for Bollinger). Use quant_trend_seasonal to strip seasonality first.

Volatility of a dated price (or return) series in annualized percent: rolling realized volatility over a window of observations or a tenor, the volatility-swap convention (log returns, zero mean, window observations), or exponentially weighted volatility of returns or of daily changes for spreads. Returns the volatility series with a line chart and its last value.

Lists every Fincept Regimes tool by module: gaussian (Gaussian, Gaussian-mixture and variational hidden Markov models), discrete (categorical, Poisson and multinomial models), inference (decode new data with a fitted model, simulate scenarios), selection (number of regimes by AIC, BIC, likelihood or holdout) and analysis (per-regime return statistics, Markov chain analytics). Give module to list one toolset.

One completion from Fincept AI over a conversation you supply; returns an OpenAI-style answer (choices[0].message.content, usage in tokens). It has no access to the platform's data, so pass the figures it should reason over in the messages. Costs 1 credit per delivered answer (action chat_message); a failed or empty answer costs nothing.

The branded model ids research_ask_model accepts, with display_name, capabilities and cost_weight.

Evaluates technical rules on the latest bar of every symbol in a universe (up to 250; the rest are dropped and reported) and returns the matches with price, bar time (ms epoch) and per-rule details, plus symbols with no data. Malformed rules are refused free with the offending path. Costs 1 credit (action screener_run), also for zero matches; not charged when no symbol could be priced.

ARIMA / seasonal SARIMA with optional exogenous regressors (ARIMAX/SARIMAX), fitted by maximum likelihood, forecasting steps periods ahead with confidence intervals (series + line chart with band). Returns the coefficient table, AIC/BIC/HQIC/log-likelihood, Ljung-Box residual checks and optional fitted values. Use it as the default forecaster for one series; stats_arma_order_select picks p and q, stats_ets and stats_theta are smoothing alternatives.

Every Fincept Stats module and its tools: regression, GLM, discrete choice, time-series models and forecasting, time-series analysis (unit roots, cointegration, decomposition, filters), multivariate time series, hypothesis tests, diagnostics, power, multivariate, nonparametric, survival, imputation and descriptive statistics. Pass module to list one. Free.

Tests a mean against a value (one sample), the mean of paired differences (paired=true), or the difference of two independent means (Welch by default, pooled Student optional), as a t or z test. Returns the statistic, p-value, degrees of freedom, the estimate with its confidence interval, the standardized effect size (Cohen's d, or Hedges' g for two samples) and each sample's n, mean and std. For equivalence use stats_equivalence_test; for three or more groups stats_anova_oneway.

Correlation matrix of the numeric columns (pairwise complete rows) with each pair's p-value. Returns the matrix as a table and a heatmap chart.

Summary statistics of each numeric column: count, missing, mean, standard deviation, standard error, 95% confidence interval of the mean, min, percentiles, max, skewness, kurtosis, Jarque-Bera normality test, mode and the share of distinct values.