Who may initiate
People, business roles, and (on the roadmap) software agents. Each gets scoped, delegated authority, never the root key.
Module 4: Policy & Automation
Available
Decide who can act, within which limits.
The policy module defines who or what may initiate a financial action, under which conditions, within which limits, and with which approvals. Every rule previews before it runs, every execution is logged, and any rule can be revoked. Today the proof rail is automation that users write in plain English (DCA, take-profit, and multi-step macros) running under scoped signing and partner-configured caps.
What policy governs.
Under which limits
Hard caps per transaction, per day, and on maximum exposure. The action cannot exceed the policy you set.
With which approvals
Require an approval step before an action runs, by amount, counterparty, or rule type. Approvals are recorded.
Every instruction the policy module governs is logged, reviewable, and revocable. Policy is one of six modules in the OVAAL control plane, and the same permissions model governs accounts and wallets. That is the foundation for programmable finance today and for agent-initiated finance on the roadmap.
The proof rail: plain-English automation.
Policy is the module. DCA, take-profit, and macros are the rail that proves it.
Schedule rules
Dollar-cost-average. Fixed amount, fixed interval, fixed asset. The most common pattern.
Trigger rules
Price, event, or portfolio-state triggered. Take-profit, stop-loss, conditional entries.
Macro rules
Multi-step actions run in sequence: for example borrow, swap, then stake.
All three use the same scoped execution engine and the same preview. A user writes a rule in plain English; the engine stores it as typed, validated, and reversible.
Rules you can defend in a compliance review.
- Preview before run. Before execution, the engine runs the full route and returns the expected outcome: expected fill, slippage, fees, and net result.
- Hard caps the user can't exceed. You configure per-user and per-transaction ceilings and a maximum exposure.
- Scoped signing. Automation runs under delegated, time-boxed keys that cannot exceed the policy, with no exposure of the root key.
- Audit trail. Every triggered rule, every execution, and every guardrail skip is logged and exportable.
- Quiet windows and kill switches. Pause all automation for a user class in one call.
Example rule.
{
"type": "trigger",
"name": "ETH take-profit",
"initiator": { "kind": "user", "id": "user_abc123" },
"trigger": {
"kind": "price",
"asset": "ETH",
"condition": ">=",
"value": 5000,
"currency": "USD",
"oracle": "chainlink"
},
"action": {
"kind": "swap",
"from": { "asset": "ETH", "amount": "10%" },
"to": "USDC",
"slippageBps": 50
},
"policy": {
"maxTxUsd": 25000,
"dailyCapUsd": 50000,
"requiresApprovalAboveUsd": 100000,
"quietWindow": { "start": "23:00", "end": "07:00", "tz": "Europe/Berlin" }
},
"expiresAt": "2026-12-31T23:59:59Z"
}
Roadmap: design partners
Ready for agent-initiated finance.
The same policy that governs human and business actions extends to software agents. An agent can initiate a financial action only inside the policies, permissions, limits, and approval rules you define. Every instruction is logged, reviewable, and revocable.
Today, policy and automation are live for human- and business-initiated flows. Fully agent-initiated finance is in design with early partners, and we are scoping it with AI-agent platforms. We mark it as roadmap rather than claiming autonomy we have not shipped.