DMNSeal
Loading secure workspace
Verifying access and retrieving the latest authorized state.
Governance infrastructure for AI agents
Human approval for high-impact agent actions.
DMNSeal gives teams a verifiable control point between agent intent and external action. Review sensitive requests, record accountable decisions, and keep execution in your own systems.
DMNSeal records decisions. It never executes the protected action.
Production / Approval request
vendor-access-2048
Grant production vendor access
- Requested by
- procurement-agent
- Protected action
- vendor.access
- Environment
- Production
- Policy
- Two-owner approval
Review progress
1 of 2 recorded
✓Operations owner approvedSecurity owner pending
DMNSeal records the decision. Execution remains external.
Synthetic product data shown for illustration.
Built for accountable operations
Clarity before consequential action.
A focused approval layer for teams adopting agents in real operational workflows.
Control
Put a person in the approval path.
Route high-impact agent requests to the right owners before another system decides whether to act.
Evidence
Keep a durable decision record.
Preserve the request, policy context, reviewers, and outcome in an integrity-protected audit trail.
Integration
Add governance without rebuilding your stack.
Create requests and retrieve decisions through a focused API designed for idempotent agent workflows.
How it works
One decision layer. Clear responsibilities.
DMNSeal handles review and evidence. Your agent and downstream systems remain responsible for execution.
- 01
Request
An agent submits a high-impact action for review.
- 02
Policy
The configured policy identifies the required reviewers.
- 03
Decision
Authorized owners approve or reject with recorded context.
- 04
Verification
The calling system retrieves the decision before proceeding.
Approval request
vendor-access-2048
Action
Grant vendor access
Policy
Two-owner approval
Requested by
procurement-agent
Approvals
0 of 2 recorded
Synthetic product data. No protected action is executed from this record.
Product evidence
Every decision has context.
Reviewers see what was requested, which policy applies, and who else must decide. Integrations receive a stable decision state and an integrity-protected record.
Built for integration
A small API surface for a critical control point.
Create an immutable approval request, then retrieve its decision state from trusted server-side code. The SDK never dispatches the protected operation.
- Idempotent requests for safe agent retries
- Stable decision states for downstream verification
- Server-side credentials designed to remain secret
create-approval-request.ts
Server onlyimport { DmnsealClient } from "@dmnseal/sdk"; const dmnseal = new DmnsealClient({ apiKey: process.env.DMNSEAL_API_KEY!,}); await dmnseal.createApprovalRequest({ actionKey: "vendor.access", requestKey: "vendor-access-2048", summary: "Grant production vendor access", payload: { vendorId: "vnd_2048" },});Private TypeScript SDK foundationCredentials never belong in browser code
Execution boundary
The decision is governed. The action stays yours.
DMNSeal preserves the request, policy context, reviewers, and outcome. Your calling system independently verifies that outcome and remains solely responsible for any external action.
- DMNSeal
- Records the governance decision
- Your system
- Retrieves and verifies the status
- External action
- Never executed by DMNSeal
DMNSeal private beta
Add accountable approval to your agent workflows.
Human approval and audit infrastructure for high-impact AI agent actions.