AI & Local Models Overview

TigerTradeBot includes an AI Copilot that can explain strategy logic, summarize backtest and optimizer results, answer questions about your trades, help draft strategies in the Strategy Designer, and review your trade journal. It’s optional: the platform works with AI turned off.

What the AI does not do

• It does not predict markets. It won’t tell you where a price is going. No model can do that reliably, and one that pretends to is dangerous.
• It does not give financial advice. It won’t tell you what to buy, when to go live, or how much to risk. Those are your decisions. TigerTradeBot is advisory software, not financial advice.

The AI and your orders

Out of the box, the AI doesn’t place orders. Everything that lets it touch trading is opt-in:

• AI Trade Confirmation Gate (off by default): in Settings → Instance Configuration → AI, AI Entry Confirmation asks the AI to confirm or veto each entry after your strategy has approved it; a veto skips the trade. AI Exit Confirmation does the same for discretionary exits (take-profit, progress and max-hold exits). Stop-loss, trailing-stop and protect-winner exits always execute immediately, and if the AI fails or times out, the exit proceeds. With the gate, the AI can only approve or veto trades your strategy proposes.
• AI Copilot chat: a trade can only start if you type CONFIRM TRADE <SYMBOL>, and then your strategy’s own rules decide whether to enter. Live mode requires an admin account.
• Webull AI Trade Intents (off by default): on the Webull page, an admin can ask the AI to draft an order for a symbol, within a symbol whitelist, a maximum order size and a daily trade cap. In manual mode you approve each one; the auto modes skip that approval. See Connecting Webull.

AI output is advisory and is never applied to your settings automatically. Parameters the AI suggests (for example from AI Optimize) have to be reviewed and saved by you.

Option A: Local models

Any OpenAI-compatible local server works. The Local LLM page has presets for LM Studio, Ollama, llama.cpp server, vLLM and LocalAI, plus a custom option. With a local model, your positions, P&L and strategy logic stay on your machine.

1. Install your runtime, download a model, and start its local server (LM Studio and Ollama both have one).
2. Open the Local LLM page, pick the preset for your runtime, and check the base URL, for example http://localhost:1234/v1 for LM Studio.
3. Set Local LLM Routing Mode (off, preferred-cloud, preferred-local or local-only) to decide how the AI chooses between your local model and any cloud provider. local-only never uses a cloud provider.

Note: Local model quality depends heavily on your hardware. Pick a model your machine runs comfortably; larger models explain code noticeably better but need more memory.

Option B: Cloud providers with your own key

Enter your own API key in Settings → Global Configuration → AI Providers: OpenAI, Gemini, Anthropic (Claude), xAI (Grok), DeepSeek or Amazon Bedrock. The AI Copilot page has a provider and model selector.

Understand the trade: the context sent with your questions, which can include strategy details and trade data, goes to that provider under its terms. Turn on AI Privacy Mode to send every AI feature whose prompts carry trade, account or position data to your local model only; it needs a configured local model. Cloud AI is opt-in, never a default.

Privacy summary

With local models, nothing leaves your machine. With cloud providers, what you send is governed by the provider’s policies. TigerTradeBot servers are never in that path.