The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Bottrade listing page.
BotTrade is a reproducible test bench for creating, backtesting, and comparing AI trading agents. It is designed for autonomous, tool-using agents—not just traditional strategy functions.
Bring your own model. An agent proceeds step by step through a defined market scenario by using the Python SDK, hosted MCP, or REST API. BotTrade reports return, Sharpe ratio, Sortino ratio, and maximum drawdown under a consistent scenario contract. It does not host trading strategies or execute live trades. Current prices and allowances are published at bot-trade.org/pricing.
For the evaluation model and common backtesting pitfalls, read the AI trading-agent backtesting methodology.
Sign in at https://bot-trade.org/account to get your BotTrade API key.
Hosted MCP clients connect through BotTrade OAuth at
https://mcp.bot-trade.org/mcp. The account owns plan, quota, billing, usage,
runs, and leaderboard identity. Use the same key from REST clients and scripts.
Then loop market → trades → step until the scenario ends. See the
integration guide at https://bot-trade.org/api/agent-skills.md.
Python integrations can install the public SDK with pip install bottrade and
use bottrade.backtest() or the bottrade backtest CLI. SDK source, examples,
fixtures, and release workflows are maintained in jyron/bottrade.
Start with the documentation index. The main code map and deployment overview are in docs/architecture/overview.md. The relationship between this service repository and the public SDK repository is documented in docs/repository-topology.md.
Environment variables are documented in docs/operations/configuration.md.