AI-Managed Vault (Design)
The AI-Managed Vault is a fully autonomous, mandate-based capital management system built on top of RAX Protocol.
It is designed to operate within explicitly defined risk boundaries, executing allocation decisions independently while remaining constrained by predefined rules and safety mechanisms.
Purpose of the AI-Managed Vault
The AI-Managed Vault exists to answer a specific question:
Can capital be managed autonomously under a clear mandate without relying on discretionary human intervention?
The vault is designed to execute decisions consistently, transparently, and within strict limits, rather than to maximize short-term returns.
Mandate-Based Design
The vault operates under a predefined mandate.
A mandate defines:
Risk tolerance and maximum acceptable risk
Target strategy profile
Exposure and concentration limits
Asset and protocol allowlists
Liquidity and volatility constraints
The AI cannot operate outside the mandate.
Mandates are explicit, auditable, and enforced at all times.
Fully Autonomous Execution
Once activated, the AI-Managed Vault:
Continuously monitors market and portfolio conditions
Evaluates risk, exposure, and liquidity in real time
Generates and executes allocation decisions
Rebalances positions as conditions change
No human approval is required for day-to-day operation within the mandate.
Risk Controls and Guardrails
Autonomy does not imply freedom.
The vault is constrained by multiple layers of control, including:
Hard risk score thresholds
Exposure and concentration caps
Liquidity minimums
Emergency stop conditions
Model confidence requirements
If constraints are breached, the vault may:
De-risk automatically
Halt reallocation
Enter a defensive or idle state
These actions are automatic and non-discretionary.
Monitoring and Transparency
All vault activity is observable.
Users can inspect:
Current mandate parameters
Allocation decisions and rationale
Risk and exposure metrics
Historical actions and outcomes
Transparency is a core requirement for autonomous operation.
Failure Handling and Degradation Modes
The vault is designed to degrade safely.
In the presence of:
Elevated uncertainty
Data degradation
Extreme market stress
The system may:
Reduce activity
Increase conservatism
Pause execution
Autonomy is reduced under uncertainty, not increased.
Governance and Mandate Updates
Mandates are not modified autonomously.
Changes to mandates require:
Explicit governance or authorized action
Clear versioning and documentation
Transition safeguards
This separation prevents silent drift in vault behavior.
Limitations and Expectations
The AI-Managed Vault does not:
Eliminate risk
Guarantee performance
React perfectly to unprecedented events
It is a constrained autonomous system, not an infallible agent.
Users must understand that autonomy increases consistency, not certainty.
Roadmap Status
The AI-Managed Vault is under active development.
Initial releases will focus on:
Narrow mandates
Conservative risk profiles
Extensive monitoring and transparency
Capabilities will expand only after sustained operational validation.
Summary
The AI-Managed Vault represents a controlled approach to autonomous capital management.
By combining full autonomy with strict mandates and hard guardrails, RAX aims to explore automation without sacrificing responsibility, transparency, or safety.
Last updated