FAQ

A trading workspace with every AI on the market, connected to your broker. You describe or pick a strategy, an agent runs it at the planned times, and you follow everything in a chat. cubrik never holds your funds.

No. You talk to cubrik like a trader on your team. Catalogue strategies are written in plain language, and yours can be described in a sentence. Before running anything, cubrik asks for what's missing: pairs, timeframe, risk per trade, leverage on futures.

In the MCP tab, add a connector: a name and the MCP URL provided by your broker or exchange, then authorise access. Under five minutes, once. Access can be revoked at any time from the same tab.

Any broker or exchange that exposes an MCP connector, across crypto, forex, indices and commodities. If yours doesn't yet, write to us: we keep a list of requests.

Yes. Seven free days in demo: real prices, simulated orders. Nothing goes live without an active plan and your explicit confirmation. We recommend at least 14 days in demo and 20 trades before moving a strategy live.

Yes. Once the strategy is launched, an agent scans the market at the planned frequency, places orders with stop and target at the broker, and sends you a message on every pass. You can ask why it didn't enter, change a setting, pause or stop it, directly in the chat.

Every hour on Starter, every 15 minutes on Pro, every 5 minutes on Elite. The scan itself is code and uses no credits; the AI is only called when a condition triggers and a decision is needed.

It stops and notifies you. Open positions keep their stop at the broker. The risk alert, which cuts everything above a max loss, keeps running: it uses no credits.

Yes, by listing several pairs at creation. The whole account capital is used: moved to futures if the strategy is futures, split 50/50 if futures and forex strategies run together.

Yes. Describe it in plain language, cubrik turns it into rules you can backtest, then run in demo. You can also duplicate a catalogue strategy and change only one rule.

Claude, GPT, Gemini, Grok, DeepSeek, Qwen, Kimi, Mistral, Llama, Muse and new versions as they ship. Three modes pick the model for you: Fast (scans, alerts), Balanced (strategies, analyses), Precise (backtests, live decisions). You can also force a specific model.

Because the right model depends on the task, not the name. An hourly scan doesn't need the most powerful model, a backtest does. Modes handle that, and your credits, for you. The list stays available for those who want to choose.

No. A strategy that loses in a range will lose in a range, whether you run it or an agent does. What AI removes is human error: the widened stop, the size doubled after a loss, the late entry, the 11pm trade out of boredom. Trading involves risk.

Never. Your funds stay at your broker. cubrik has an access you authorise, to read your accounts and place orders, which you can cut at any time.

Stops and targets are placed on your broker account, not kept in the agent's memory. They hold even if our servers are down. No new order is placed in the meantime.

Three, independent: the max daily loss on every task, the hourly risk alert that cuts everything above a global threshold, and the stops at the broker. None depends on the AI or your credits.

Your broker access tokens are encrypted. No data is used to train models. You can delete your history at any time. GDPR compliant.

Each plan includes monthly AI credits. Fast mode uses few, Precise mode more. You track usage in Settings › Usage. Extra credits can be enabled with a monthly cap you set.

At any time, from the web version. Upgrades are immediate; downgrades or cancellation apply at the next renewal.

Yes, on iOS, with the same conversation and the same agents as on the web. Plans are managed on the web version.