Current trading configuration (secrets redacted). The single source of truth for mode, permissions and risk limits.

Effective trading mode. 'live' only when TRADING_MODE=live AND ALLOW_LIVE_TRADING=true.

Update runtime-updatable config keys in .env (risk limits, intraday/swing enable flags). Live-trading switches, ALLOW_DELIVERY_SELL and credentials can NOT be changed here — a human must edit .env.

Set TRADING_MODE=paper (always safe).

Set TRADING_MODE=live. NOTE: real orders remain impossible until a human also sets ALLOW_LIVE_TRADING=true in .env.

Check config permissions for a prospective order (no market/risk checks). side: BUY|SELL, product_type: INTRADAY|DELIVERY.

Last traded price for an NSE symbol (requires broker credentials).

Historical OHLCV candles. timeframe: ONE_MINUTE|FIVE_MINUTE|FIFTEEN_MINUTE| THIRTY_MINUTE|ONE_HOUR|ONE_DAY. Dates ISO format (optional).

Whether the market is open, pre-open or closed right now (IST).

Instrument metadata: token, exchange, lot size, tick size.

Watchlist symbols: Chartink monthly-uptrend + RSI>50 scan, falling back to a static NIFTY list when the scan is unavailable.

Simple moving average (latest value).

Exponential moving average (latest value).

Wilder RSI (latest value).

MACD line, signal and histogram (latest values).

Bollinger bands (latest values).

Average True Range — useful for stop-loss placement.

Volume vs its EMA, spike detection, 20-bar average.

Support/resistance levels from clustered swing pivots.

Trend direction from EMA stack + regression slope.

All standard indicators in one call (RSI, EMAs, MACD, Bollinger, ATR, volume, trend, VWAP on intraday timeframes). Prefer this over many single calls.

Latest news items for one symbol with per-item sentiment.

Broad market news. category: all|stock|sectoral|global; sentiment: all|positive|negative|neutral.

Aggregate sentiment counts for a symbol's recent news. The agent should read the texts (fetch_latest_news) for nuance — this is just the tally.

Portfolio for the EFFECTIVE mode: paper portfolio in paper mode, broker holdings+positions in live mode.

Order history for the effective mode (paper orders, or broker order book).

Per-position exposure vs MAX_POSITION_SIZE_PERCENT (paper portfolio).

Paper portfolio P&L. Pass live_prices as {symbol: ltp} (e.g. from fetch_live_price) to include unrealized P&L.

Quantity so the stop-loss loss equals risk_percent of capital, capped by MAX_POSITION_SIZE_PERCENT.

ATR-based stop-loss suggestion from current price.

Target price for a desired risk:reward given entry and stop.

Whether today's realized paper P&L has hit MAX_DAILY_LOSS_PERCENT.

Concentration check across open paper positions.

Full pre-trade validation checklist: permissions, market status, risk limits, stop/target presence, risk:reward, position size, daily loss, open positions. Run this BEFORE any paper or live order.

Multi-timeframe (daily/1h/15m) intraday setup with a mechanical trade plan and confidence score.

Daily-timeframe swing/delivery setup with support/resistance and a mechanical trade plan.

Side-by-side indicator snapshots for up to 10 symbols.

Evaluate the watchlist for swing setups, ranked by confidence score.

Evaluate the watchlist for intraday setups, ranked by confidence score. Slow: fetches three timeframes per symbol.

Backtest a built-in strategy (ma_crossover | rsi_reversal | macd_trend | breakout_volume) on historical data. Returns trades, win rate, P&L, drawdown.

Place a SIMULATED order in the paper account. If price is omitted the live LTP is used (requires broker credentials). The full validation checklist is enforced here too.

Close an open paper position at the given price (or live LTP).

Closed paper trades, optionally filtered by INTRADAY or DELIVERY.

Paper account: cash, open positions, realized P&L.

Performance metrics: win rate, P&L, drawdown — overall, per product type, and per strategy. Use this to judge profitability before considering live mode.

Full paper-trading report (performance + portfolio + recent trades).

Reset the paper account to starting capital, erasing simulated history. Requires confirm=true.

Validate and STAGE an order. Never executes. In live mode returns an approval_token; the human must approve before execution. Delivery sells are blocked and returned as recommendation-only.

Re-run the full validation checklist for an order without staging it.

Execute a prepared INTRADAY live order. Only call after the human has explicitly approved the prepared order shown to them.

Execute a prepared DELIVERY BUY live order. Only call after the human has explicitly approved. Delivery SELL can never be executed through this system.

Record a delivery-sell RECOMMENDATION. The order is never sent to the broker — delivery sells require manual verification by the user.

Prepared live orders awaiting human approval.

Cancel a prepared live order before execution.

Available funds/margins from the broker (requires credentials).

Today's positions from the broker.

Delivery holdings from the broker.

Status of a broker order by id.

Send a free-text message to the configured Telegram chat.

Requires TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID in .env; if either is missing, returns {'sent': False, 'reason': 'telegram_not_configured'} without raising. parse_mode: HTML | Markdown | none. This sends a MESSAGE ONLY and never places an order.

Format an 'ACTION REQUIRED' trade recommendation and send it to Telegram for the human to place manually (used in notify-only mode).

side: BUY | SELL; trade_type: intraday | delivery; mode: paper | live. Sends a RECOMMENDATION ONLY — it does NOT place any order. The returned dict always includes 'preview' (the message text) so the agent can also surface it inline even when Telegram is not configured.