stratify-mcp

Python client for Stratify — real 1-minute NIFTY options data, honest backtests (out-of-sample, walk-forward, deflated Sharpe, all reported alongside the number, not instead of it).

pip install stratify-mcp

Quickstart

from stratify_mcp import StratifyClient

# One-time: create an account and get a key. The key is shown once -- save it.
signup = StratifyClient.signup("you@example.com")
client = StratifyClient(api_key=signup["api_key"])

result = client.run_backtest({
    "legs": [
        {"side": "sell", "type": "CE", "strike": {"delta_near": 0.2}},
        {"side": "sell", "type": "PE", "strike": {"delta_near": 0.2}},
    ],
    # The reason for the trade, not just the trade.
    "entry": {"cadence": "weekly", "dte": 3, "time": "09:30",
              "when": {"vix": {"gte": 15}}},
    # Managed while it is open: take profit, and roll the tested side if it doubles.
    "rules": [
        {"when": {"pnl_pct_of_credit": {"gte": 0.6}}, "then": "close"},
        {"when": {"leg_mark_mult": {"gte": 2.0, "leg": 0}},
         "then": {"roll": {"legs": [0], "to": {"delta_near": 0.2}}}, "max_times": 2},
    ],
    "portfolio": {"stop_after_losses": 3, "resume_after_days": 30},
})

print(result.summary["total_pnl_rupees"], result.summary["max_drawdown_rupees"])
print(result.summary["ratios"])       # sharpe, profit_factor, calmar (only above 30 trades)
print(result.honesty)          # out-of-sample split, walk-forward folds, deflated Sharpe
result.trades                  # pandas.DataFrame, one row per trade
result.equity_curve            # pandas.DataFrame
print(result.report_url)       # shareable page with the full chart and every trade

Already have a key? Skip signup():

client = StratifyClient(api_key="sk_live_...")

Why a Python client at all, when it's just JSON-RPC

There's no separate REST endpoint for run_backtest — every tool is reached through one POST /mcp speaking MCP JSON-RPC 2.0. This package exists so you don't hand-roll that envelope: client.run_backtest(...) is a real function call, errors come back as Python exceptions you can except, and results come back as pandas.DataFrames instead of raw JSON, because that's what you're actually going to do with a table of trades.

Errors

from stratify_mcp import AuthenticationError, QuotaExceededError, ToolRefusalError

try:
    result = client.run_backtest(spec)
except AuthenticationError:
    ...  # bad or revoked key
except QuotaExceededError as e:
    ...  # e.limit, e.retry_after_seconds
except ToolRefusalError as e:
    ...  # the server read your spec and refused it -- str(e) says why

Methods

detail on run_backtest/get_backtest: "summary" (aggregates only, cheapest), "standard" (default — equity curve, breakdowns, first 25 trades), "full" (every stored trade).

BacktestResult

Development

pip install -e ".[dev]"
pytest

Tests run against a mocked transport and need no live server or API key.

License

MIT.