Answer five questions
Start with the strategy guide. Set your goal, simulated allocation, acceptable loss, research pace and decision mode. Observation, per-action confirmation and automatic simulation have different authority boundaries; choose the one that reflects your intent.
Review the proposal and constraints
Self-selected strategies and AI recommendations enter the same configuration flow: choose assets and an account, review parameters, allocation and protection, then check and approve the current version. Closing configuration returns to its research or strategy source. Saving a plan does not start a run; strategy drafts and running instances remain separate.
Turn the research into a reviewable plan
An AI BOT approval covers a specific plan. Changes to the account, material parameters or execution scope require another check and approval. Autonomous AI trading is designed to choose opportunities within an explicitly authorized scope; its independent authorization and continuous operation remain under development. Signing in, connecting, reading data and executing trades are distinct permissions.
Review the records
Pausing new trades does not sell existing holdings. Ending a bot requires an explicit choice to retain holdings or request an exit; actual receipts establish the outcome. Partial fills, unconfirmed orders and residual holdings still need reconciliation. Query the original operation when its result is unknown rather than treating a timeout as failure and submitting again.
Current evidence covers adapted grid and bounded DCA versions in local simulation. Catalog inclusion does not establish executability, and joint acceptance of two BOT flows is still pending. Simulation records cannot prove real fills or a live trading connection.