Our ambition is to make agent trading infrastructure that people can control. The person supplying the capital should decide who may act, in which markets, and within what limits. Delegating execution should preserve that authority, including the ability to withdraw permission.
An agent should be free to reason within the authority it has been given. The limits themselves belong outside its discretion. We are building an execution layer that evaluates requests before signing and returns a clear refusal when they exceed the operator's policy.
Control means more than holding a wallet. It means choosing the agent, retaining the account-owner keys, understanding where capital is allocated, and keeping a record of what happened. Oppen's local-first architecture is designed to put those decisions within the operator's reach.
Delegation creates a responsibility to account for outcomes. An operator should be able to distinguish what an agent requested, what the system permitted, and what the venue actually executed. Uncertainty, refusals and failures belong in that record as clearly as successful trades.
Access to programmable markets should come with the ability to examine the software acting on them. We intend to publish the enforcement core under Apache-2.0 at launch. Trust should rest on inspectable mechanisms and demonstrated behavior, with the limits of each stated plainly.
More activity is not progress by itself. A system should demonstrate that it respects permissions, survives interruptions and responds to its operator before it receives greater authority. Our starting point is one venue, testnet and measured execution. Broader capability follows proof.
The name oppen reflects a conviction that powerful tools deserve deliberate stewardship. Our contribution is specific: make delegated trading easier to understand, constrain and supervise. We will judge the work by the control it gives its users and the evidence behind its claims.