Summary
This guide walks through building a free AI trading bot using DeepSeek as the model, OpenCode as the agent harness, and the TradingKit MCP server for strategy backtesting. OpenCode is presented as a free alternative to Claude Code, and the video shows how to install it, switch to a free DeepSeek model with the /models command, and connect it to TradingKit without needing a TradingView account.
The core of the tutorial is an autonomous research loop. A single long prompt casts the agent as a quantitative trading researcher that tests and optimizes strategies every 15 minutes for 24 hours, roughly 96 cycles. It includes guardrails against overfitting and filters out strategies with fewer than 100 trades or large maximum drawdowns. The video also tours TradingKit’s control panel, where backtest results, equity curves, and trade lists can be reviewed.
The creator stresses caution: backtests that look too good are usually unreliable, so promising strategies should go through a one-to-three month paper-trading incubation period before any live money is used. The final step connects a chosen strategy to a broker through the trigger.trade automation service. Viewers interested in agent-driven workflows, MCP tooling, and low-cost AI setups will find the full prompt and tool sequence useful.
📺 Source: Trading with DaviddTech · Published October 19, 2026
🏷️ Format: Tutorial Demo
