Customer control
Why do risk limits and exact plan consent matter?
Understand what a trading plan must clarify, why consent belongs to one version, and why registration cannot activate orders.
Risk limits define the boundaries an execution system must respect. Exact plan consent records which proposal the customer reviewed, rather than a general permission for future changes. In TradeTwin, acceptance, account readiness and execution activation are separate gates. A draft, an interview answer or a workspace request is not an order authorization.
Specify the boundaries before automating decisions
A plan needs more than an entry idea. It should identify sizing, maximum exposure, the number of simultaneous positions, reserved costs and the protection the venue can actually maintain. A limit needs a clear meaning and unit: a maximum lot size is different from a fraction of account equity. Missing inputs should remain visible. AI confidence cannot substitute for a saved limit or resolve an ambiguous risk instruction without the customer's review.
Accept one proposal, not every later revision
The private proposal flow binds acceptance to the proposal's version and content hash. This identifies the exact plan that was reviewed. If new information leads to another proposal, earlier acceptance is invalidated for that changed plan. This separation helps prevent a helpful-looking suggestion from becoming an unnoticed instruction. The customer should inspect altered assumptions and risk fields, not simply assume a familiar plan title means that all its details stayed the same.
What does acceptance actually do?
Acceptance records a decision about the proposed plan; it does not arm trading by itself. The selected account must be verified, supported rules must represent the strategy, and native protection must be available for the intended execution path. No external MCP tool can provide the owner's consent or submit an order in the current private proposal interface. This avoids confusing approval of written reasoning with permission for a particular broker action.
Separate stopping the interview from managing positions
A customer can stop answering questions or choose Enough to request a draft. Those controls affect the interview, not an open market position. Likewise, stopping a pending AI response does not prove the provider never processed it or charged for it. An execution stop, a new-entry pause and a position close have different consequences. The exact controls and the remaining protective orders must be verified in the activated customer's own workspace.
Use outcome evidence without expanding limits silently
A series of wins can justify further investigation, but it does not justify automatically increasing size or bypassing a maximum exposure boundary. A loss can reveal a rule ambiguity, a cost problem or a market condition; it should be reviewed with entry-time evidence. Compare proposals before accepting an adjustment. Useful learning keeps the customer's saved limits intact unless they explicitly approve a supported change, and it does not treat a target win rate as an observed result.
Questions worth asking
Does accepting a plan start an order?
No. Plan acceptance is separate from execution activation. Account verification, representable rules and the venue's required protection must pass their own checks before an order path can be enabled.
Can a new plan reuse my old consent?
A replacement proposal must be reviewed again. The existing private flow binds consent to the exact version and hash, and invalidates earlier consent when a newer proposal is created.
Can I request more positions?
You can express that objective, but it does not override saved risk limits or prove qualifying entries exist. Any supported exposure change needs review of the exact plan and account-specific capacity; an objective is not a performance result.