README.md

# Premiss

Turn your next trading idea into a strategy you can test and improve, right from ChatGPT or Claude.

Explore market data with tables and charts, then describe your rules in plain English. Create custom Python strategies, backtest supported crypto, stock, and ETF markets, and see the trades behind the results. Explore drawdowns, inspect equity curves, and compare completed tests with matched buy-and-hold evidence.

Your strategies, tested code, and recorded results stay in your private Premiss workspace. Pick up an earlier idea, refine the rules, and run it again. Read your paper simulations to see which version is running and what it last did.

Connect your Premiss account and choose a workspace and permissions. Editing existing strategies needs separate edit consent; paper evidence needs separate read consent. Custom indicators, entry/exit logic, risk rules, and long/short/flat strategies use the app's package format and sandbox. The app's market availability, code validation, and active-run protections apply. Backtests default to 10,000 simulated capital and modeled next-eligible-open market fills. New backtests default to zero fees and slippage. You can ask for specific fees and slippage. Every result records the assumptions actually used.

Strategies can watch several markets and timeframes at once: one market can signal a trade in another, a daily trend can filter hourly entries, and rules can run on a schedule, after fills, or with partial exits. A single entry uses the available capital unless you choose a size; simultaneous entries need explicit allocations. Portfolio strategies can rebalance to target weights, such as 60% BTC and 40% ETH, and keep part in cash. Prices, fees and quantity steps can make the achieved split differ slightly from the target. Stock and ETF prices are not adjusted for splits or dividends, and every stock result says so. Multi-market results show the full portfolio and do not invent a buy-and-hold benchmark.

When a test looks worth following, Premiss can set up a paper bot that keeps the tested markets, parameters, capital and costs. Stock and ETF paper bots check completed daily or weekly candles.

Premiss supplies market data analysis, the engine, saved versions, recorded trades, result charts, and ongoing paper execution. ChatGPT or Claude handles the conversation, code authoring, explanations, and any documents you request. A completed backtest does not automatically start a bot; paper setup and controls remain in the Premiss app.

Historical results are not forecasts. The connection does not place live trades, control bots, expose exchange keys, change billing, or publish content.

Support: https://app.premissai.com/support. Connection instructions: https://app.premissai.com/integrations. Manage or revoke connections: https://app.premissai.com/integrations/connections. Privacy: https://premissai.com/legal/privacy. Terms: https://premissai.com/legal/terms.

The repository contains manifests for both hosts. Packaging produces separate ZIPs with one host's MCP resource and the shared strategy-testing skill. Hosted Claude uses its web callback; Claude Code's local OAuth callback is not supported. Public directory availability requires each host's review.