Ask, and it researches
Candles, indicators, order flow, funding and open interest — pulled live and read back as an argument you can disagree with, not a verdict.
Opening 1 December 2026 · paper accounts first
Trader Console researches markets, screens for setups and backtests strategies in one conversation. Every order it wants to place is checked against limits you set — and it has no way to raise them.
Proposed Open 1.4% of equity, stop 3.1% below entry
Every account starts in paper. Live keys are added by you, and the console can trade only the accounts you connect.
The parts a retail trader normally runs across five tabs and a spreadsheet, joined by a system that remembers what it decided and why.
Candles, indicators, order flow, funding and open interest — pulled live and read back as an argument you can disagree with, not a verdict.
Set the conditions once; they run on a schedule and tell you when something crosses. Every firing is kept, so you can see what the screen actually caught.
Walk-forward runs with a leakage check, because a strategy that has read the future backtests beautifully and loses money.
Nothing goes live on a good backtest alone. A strategy trades on paper first, and the promotion to real money is a decision you sign.
paper → live
Position size, daily loss, exposure and leverage are evaluated by code the model cannot call, edit or talk around.
every order
Every fill traces back to the decision that caused it — the question you asked, the proposal, the checks that ran, the order id at the venue.
Four steps, and the boundary sits in the middle of them. This is the whole security argument of the product, so it is on the front page rather than in a whitepaper.
In plain language: find me a setup, size this position, tell me what changed in my book overnight.
It reads market data and your positions, and produces a proposal with its reasoning attached. A proposal is text. It moves nothing.
A separate service evaluates the proposal against your limits and signs the result. It runs on rules, not on judgement, and the model has no route to it other than being checked by it.
Live orders wait for your confirmation. Once you allow automatic execution for a strategy, the gate still runs on every single order — that permission widens what may pass, never what is checked.
A capability list is easy to write. This is the other one — the things the architecture makes impossible, which is what actually protects an account.
Only if you turn that on, per strategy, after it has traded on paper. By default it proposes and waits. Either way every order is checked by the risk gate first — automation changes who confirms, not whether the check runs.
In a secrets store keyed to your account, and they are used by the execution service alone. They are never handed to the model, never printed into a chat, and never shared with another customer’s process.
You set a daily loss limit. When the account reaches it, the gate stops approving new entries for the rest of the day — including the model’s. Reducing or closing a position stays allowed, because blocking an exit is how a bad day becomes a bad month.
Yes, and this is the part most tools skip. Each order carries the decision that produced it: what you asked, what the model proposed, which checks ran, what the venue answered.
No. There is no basket of tips and no copy trading. It is your account, your limits and your confirmations — the model is research and drafting, not a manager.
A paper account runs the whole system — the same research, the same screeners, the same risk gate — against live market data and no money.