# AlphaProve

> AlphaProve (alphaprove.com) is a crypto backtesting platform for algorithmic
> trading. Users describe a trading strategy in plain English (an AI assistant
> writes the Python), write it themselves, or use a no-code builder — then
> backtest it on years of 1-minute crypto market data with L2 order-book fills,
> validate it with forward-walk (walk-forward) analysis, and judge it by
> Sharpe ratio, drawdown, and net PnL from a full tearsheet.

Key facts:

- Backtests run server-side in a sandboxed engine; fills are simulated against
  real L2 order-book depth when coverage exists, with a candle-based fallback.
- Forward-walk validation and Monte-Carlo risk simulation separate robust
  strategies from curve-fit ones.
- Strategies are written against a frozen, versioned Python SDK contract
  (`strategy_sdk.v1`) with an `evaluate(ctx)` entrypoint.
- Tiers: Free, Trader, and Quant. The Python editor is available from the
  Trader tier; free users can view AI-generated code read-only.

## Connect an AI agent

- [Homepage](https://alphaprove.com/): public product overview.
- [MCP setup guide](https://alphaprove.com/mcp): connect Claude, ChatGPT,
  Codex, or another compatible client.
- [MCP protocol documentation](https://alphaprove.com/docs/mcp/): setup,
  authentication, scopes, and available tools.
- MCP endpoint: `https://alphaprove.com/api/v1/mcp` (Streamable HTTP).
  Account tools require OAuth and an approved beta account. Public pages and
  documentation can be read without signing in. A bare browser GET to the
  protocol endpoint is not the setup guide.
- [OAuth resource metadata](https://alphaprove.com/.well-known/oauth-protected-resource/api/v1/mcp)
- [OAuth authorization server metadata](https://alphaprove.com/.well-known/oauth-authorization-server/api/v1/oauth)

## Official profiles

- [LinkedIn](https://www.linkedin.com/company/alphaprove)
- [X](https://x.com/Alpha_Prove)

## Docs

- [Getting started](https://alphaprove.com/docs/getting-started/): how to go
  from a plain-English idea to a backtest and tearsheet.
- [Strategy authoring](https://alphaprove.com/docs/strategy-authoring/): the
  Python `evaluate(ctx)` contract, indicators, signals, and parameter sweeps.
- [Risk management](https://alphaprove.com/docs/risk-management/): sizing,
  leverage, stops, and engine-enforced limits.
- [Examples](https://alphaprove.com/docs/examples/): complete strategies ready
  to backtest.
- [SDK reference](https://alphaprove.com/docs/sdk-reference/): the published
  `strategy_sdk.v1` contract.

## Blog

- [We tested ChatGPT's trading strategies](https://alphaprove.com/blog/we-tested-chatgpts-trading-strategies):
  four ChatGPT-proposed strategies (MACD+ADX, EMA crossover, time-series
  momentum, Donchian breakout) implemented unmodified and backtested on two
  years of BTCUSDT 1-hour data. All four lost money (-7% to -44%) while
  buy-and-hold gained ~15%. The case for using AI for implementation rather
  than idea generation.

## Glossary

- [Trading glossary](https://alphaprove.com/glossary): plain-English
  definitions of trading, backtesting, and quant terms (long/short, Sharpe
  ratio, forward-walk analysis, futures basis, market structure, and more) —
  one page per term.

## Legal

- [Terms of Service](https://alphaprove.com/legal/terms)
- [Privacy Policy](https://alphaprove.com/legal/privacy)
- [Risk Disclosure](https://alphaprove.com/legal/risk-disclosure): backtests
  are simulations; past performance does not guarantee future results.