Mandate Firewall
Verify what an agent is allowed to do before execution. Scope, caps, counterparties, and policy rules are enforced server-side before any value moves.
- Per-agent mandates
- Counterparty classification
- Daily caps and policy guards
AgentGuard · Control Plane
AgentGuard Control Plane
The black box, mandate firewall, and settlement workflow layer for AI agents that touch money.
Payment protocols move money for AI agents. AgentGuard verifies intent, detects mandate violations, preserves evidence, and routes financial failures into auditable settlement workflows.
Three pillars
Mandate. Evidence. Settlement.
Action Black Box
Capture intent, payload, decision hash, and receipt for every authorized action. Hash-chained audit events make every step independently verifiable.
- Authorization records
- Execution receipts
- Hash-chained audit events
Settlement Workflow
When an action goes wrong, AgentGuard opens an incident case, verifies evidence, and routes it into partner-approved settlement instructions.
- Incident verification
- Multi-party approval
- Partner-approved settlement
Why now
Agentic payments are arriving faster than agent oversight.
AI agents are starting to call payment protocols, x402-style flows, stablecoin transfers, partner payouts, and wallet actions on behalf of organizations. Most of those actions ship with no mandate verification, no consistent audit trail, and no settlement path when something goes wrong.
AgentGuard sits in front of those actions: it checks the mandate, evaluates policy, records hash-chained evidence, and gives partners a structured way to verify incidents and approve settlements.
Architecture
An extension of Cryptodam's incident protection architecture.
AgentGuard extends Cryptodam's digital-asset incident protection architecture into AI-agent-driven financial actions. The same modules that govern monitoring, trigger verification, eligibility, and payout workflows are reused to verify agent mandates, classify action attempts, and route verified incidents into partner-approved settlement.
- Policy enforcement
- Mandate verification
- Incident verification
- Settlement workflow
- Payout instruction
- Partner-approved settlement
Integration
Partner endpoints.
Partners integrate AgentGuard through scoped API endpoints. Credentials and scopes are managed in the Partner Console — no secrets are exposed on this page.
| Method | Path | Purpose |
|---|---|---|
| POST | /functions/v1/agentguard-authorize-action | Evaluate a proposed action against mandates and policy. |
| POST | /functions/v1/agentguard-record-action-receipt | Record execution payload and produce an evidence trail. |
| POST | /functions/v1/agentguard-x402-webhook | Inbound x402-style payment-protocol receipts. |
Product boundary
What AgentGuard is — and is not.
AgentGuard is
- Policy enforcement for agent actions
- Mandate verification before execution
- Incident verification with hash-chained evidence
- Settlement workflow into partner-approved resolution
AgentGuard is not
- Licensed insurance
- A guaranteed-payout product
- An underwriting service
- A custody or wallet provider