Render a go/no-go before an agent moves money
Before an autonomous agent executes a payment, FX conversion, treasury sweep, or drawdown, LyraMind renders an execution-readiness decision. It composes an intelligence gate, a notional policy, and provider routing into one recorded verdict. LyraMind decides and orchestrates; it never moves funds.
The problem
An agent that can execute a financial action needs a control in front of it: is the underlying intelligence grounded and fresh, does the size demand human confirmation or dual control, and is there even a connected provider that could execute it? Without that gate, an agent can act on stale or ungrounded data, or move a notional that should have required two approvers.
How LyraMind does it
Call the execution decision with an action (type, capability, amount, currency, counterparty) and an optional symbol. LyraMind reuses the Agent Readiness gate for the intelligence basis, escalates on stale data, applies a deterministic notional policy (an auto-execute limit and a dual-control threshold), and routes to the connected providers that could execute the capability. The decision is fail-safe: any single blocking condition holds, and policy can only escalate toward caution.
What you get
A single decision object: clear-to-execute or not, why, the controls applied (confirmation required, dual control, stale intelligence, no provider), the eligible providers, and a ledger id so the outcome can be reconciled later. It is a control and gate, never a buy/sell recommendation.
- Pass an action: type, capability, amount, currency, counterparty
- Intelligence gate grounds it when a symbol is supplied
- Notional policy applies auto-execute and dual-control thresholds
- Provider routing lists who could execute the capability
- Verdict: PROCEED / READY / WAIT / REVIEW / NOT_READY / HIGH_RISK
- Decision, reasons, and accountable actor recorded to the Trust Ledger
Questions
Does LyraMind execute the trade or move the money?
No, and this is load-bearing. LyraMind renders the decision and routes to connected providers; it never moves funds, custodies assets, or executes. Connecting a provider is what enables execution, performed by that provider.
What happens with missing inputs?
The decision degrades honestly rather than guessing. No amount or a non-numeric amount escalates to WAIT; no connected provider for the capability escalates to NOT_READY. It is fail-safe by design.
Are the providers real?
In the demo they are integration stubs, with WireUp as a first-class partner slot. The seam is real; a live provider integration fills it. LyraMind still never executes.