No purchase can be the correct outcome of a spending task. An agent may lack enough information, reach a permission boundary, or encounter an unresolved transaction. The important thing is whether it tells you which situation occurred.
A system that always produces a success message is not necessarily more capable. It may simply be hiding the difference between research, permission, and execution. A useful stop gives you enough information to decide what, if anything, should happen next.
The research may be incomplete
An unavailable source, a missing token identifier, or conflicting information can prevent a meaningful selection. The agent can return what it knows without pretending the missing piece is established.
If you asked for live X research, a prepared example is not an equivalent replacement unless it is clearly labeled. It can demonstrate the experience, but it cannot answer what people are discussing at this moment.
The Veyro product page distinguishes fixtures from real X results. When inspecting the demo, use that source information to understand what kind of result you received.
The action may exceed permission
A candidate can look relevant while the proposed action remains outside the allowed spend. The task may also have expired or its authority may have been revoked. Those are reasons to stop, not reasons for the agent to invent a different allowance.
Start by reading the explanation. If the amount exceeds the remaining budget, ask whether the action is still desirable at that amount. Do not increase the limit merely to get a green result.
Read agent spending limits and permission expiry if the message concerns the boundary rather than the search.
The requested action may not be supported
A product can support a narrow demonstration without supporting every token, route, or purchase a prompt can describe. Natural-language input is broad; the available actions may be much narrower.
If the agent cannot carry out the request, the helpful response is to name the limitation. It should not silently choose a different network or unrelated destination to make the task appear complete.
This matters when moving from an example prompt to a custom one. The ability to type a request does not establish that the system supports every action mentioned in it.
A network response can remain unresolved
A slow response or interrupted connection is different from a rejected action. The transaction may already have been submitted. Repeating the request can make it harder to tell which attempt produced which result.
Check the existing task status and any transaction reference first. If the outcome is unknown, preserve that uncertainty. A product should not report “nothing happened” merely because it stopped waiting for a response.
See transaction receipts for how to connect an attempt to network evidence.
Decide what kind of next step is needed
Missing information may call for a clearer prompt. A permission issue may call for a deliberate review of the task. An unsupported action may mean the product cannot do it. An unresolved transaction calls for status checking before another attempt.
These situations should not all lead to the same instruction to try again. The next step should address the actual reason the agent stopped.
The Veyro demo is built around a task, a limit, and a visible decision. Judge that decision by whether it respects the task you gave it. Sometimes the most valuable thing an agent can do is explain why it did not spend.
Your task. Your limit.
Explore the current Veyro experience, or get to know the product first.