Vault Agent preferences

Customization builds up in layers: an investment profile sets sensible defaults, natural-language preferences refine or override those defaults for anything specific to you, and the system stores what you say as memories that compile into the rules the Allocation Engine actually applies. Almost anything a profile sets can be adjusted through natural language — a profile is a starting point, not a ceiling.

Investment profiles: the defaults

The Vault Agent currently offers two profiles:

Conservative (the default) — sticks to established, overcollateralized vaults with a proven track record.
Explorative — targets higher yields across a wider range of vaults.

A profile sets the starting weighting the Allocation Engine applies during allocation and continuous monitoring, including how much it favors newer or higher-yield opportunities over established ones. Choosing a profile does not grant access to unsupported protocols or remove platform safety checks, and names or availability can change as the strategy evolves.

Natural-language preferences: refine or override the defaults

Use the agent’s Discuss or Steer experience to describe a preference in plain language. You can ask it to prefer a protocol, avoid a vault, prioritize more stable opportunities, or tighten a limit your profile sets more loosely — a preference can narrow, bias, or override most of what a profile decides by default.

A preference is not an arbitrary transaction instruction. The agent interprets it as an allocation preference and applies it only where the resulting behavior is supported; it cannot make an unsupported or unsafe target eligible.

Memories and rules: how preferences persist

This is the storage layer underneath both of the above. The system saves an accepted preference as a memory and compiles it into one or more rules used by future Vault Agent decisions alongside your chosen profile. Review the saved memory and its active status in the app.

Rules can interact with one another. A new instruction may replace or constrain an existing rule rather than simply adding to it, and can just as easily narrow something your profile would otherwise allow.

Simulate before saving

When the app offers simulation, use it to preview how a preference could change candidate scoring and target allocations. A simulation is a projection, not a transaction, and does not guarantee a future result.

Ask why

You can ask the agent why it selected a vault, stayed in a position, or did not follow an apparent preference. Explanations describe the available decision context; they are not a guarantee that every future decision will be identical.