Cryptodamby StuporeAgentGuard · Control Plane
The black box, mandate firewall, and settlement workflow layer for AI agents that touch money.
AgentGuard checks proposed actions against recorded mandates and policy, preserves action records, and routes incidents through an auditable settlement-review workflow.
Authority before action
AgentGuard gives a partner’s systems a place to evaluate an agent’s proposed action before execution. Connect the agent identity and recorded mandate to policy rules, a decision with reasons, and a receipt describing what actually happened.
Who it helpsAI product teams, platform operators, partner risk teams and compliance reviewers.
Explore the process
Agent identity, mandate and a proposed action
An approval, escalation or block with a reviewable record
An example journey
An agent requests an action within its amount limit, but the counterparty needs review. The request can be escalated with the relevant policy reason, giving a person the context to decide before the partner’s system proceeds.
Three pillars
Requests submitted to AgentGuard are checked server-side against recorded scope, limits, counterparties and policy. The partner system must enforce the returned decision before proceeding.
Record the supplied action payload, authorization decision and received receipt. Integrity hashes link related audit events for authorized review.
When an action goes wrong, AgentGuard opens an incident case, verifies evidence, and routes it into partner-approved settlement instructions.
Why now
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
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.
Integration
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
AgentGuard is
AgentGuard is not