About
A trading console for exploring AI judgment while keeping execution under boring, explicit rules.
LLM Trader is a trading lab built around a simple rule: the AI can have ideas, but it does not get to touch the keys. Local language models look at the market and suggest SOL/USDC swaps, then a separate risk layer decides whether the idea is even allowed.
The useful part is the separation. The system checks prices again at submit time, applies the slippage and size limits, and rejects anything that has drifted too far. The models can be interesting without being trusted.
Most of the work lives in paper mode, where the whole path can run without moving real money. That makes it a place to study behavior, not a black box pointed at a wallet.
Highlights
- Local models can propose trades, but a separate rule layer has the final say.
- Paper mode lets the full flow run without putting funds at risk.
- The dashboard keeps the important state visible: balances, trade state, risk settings, and provider health.
- Wallet details are blurred in portfolio media so the screenshots show the product, not sensitive identifiers.
Also worth noting
- The current interface covers the dashboard, wallet, positions, risk policy, and system health views.
- The system is designed to be inspected before it is trusted.
- The screenshots are intentionally sanitized before being shown publicly.