Authorize dangerous actions before agents execute.
SilentAuth gives AI agents and automations a receipt they must verify before acting. Policies handle routine cases; humans and local proof handle the rest.
Protect an action
Create a project, register an agent, and decide which actions must stop for policy or human review.
Approve and issue
Resolve the request with policy, a reviewer, or quorum, then issue a short-lived signed receipt.
Require local proof
For the highest-risk actions, route approval through a Mac Mini gateway with VINAC-FM, QR, or NFC proof.
Classify intent
Use Logos Lexicon later to describe bounded machine intent without passing arbitrary instructions.
Built around the action, not the chat.
The first version should do one thing clearly: stop risky automation, resolve approval, issue a receipt, and leave an audit trail.
Project-scoped agents
Give every automation a scoped identity before it can request permission for production work.
Policy decisions
Route each intent to auto-approve, auto-deny, single reviewer, quorum, or local proof.
Receipts agents verify
Issue short-lived receipts that bind action hash, policy hash, nonce, approver, and expiry.
Local proof gateway
Use the hosted control plane first, then connect a Mac Mini gateway for VINAC-FM, QR, and NFC approval proofs.
Logos Lexicon
Classify what an agent wants to do with bounded intent tokens that policies can evaluate consistently.
Operations hooks
Send approval, denial, receipt, and proof events to the systems your team already watches.
Who needs action authorization?
AI agent builders
Give agents a controlled way to ask for permission before sending email, changing records, deploying code, or touching customer data.
Register an agentPlatform teams
Route CI, deployment, Terraform, and internal automation through approvals that can be verified before execution.
See execution gatesSecurity reviewers
Review what happened, who approved it, which policy applied, and which proof channel was required for sensitive actions.
View audit modelOne loop every protected action follows
Account, project, agent, policy, approval, receipt, audit. Keep the first experience this concrete.
Register the actor
Start with project scope.
A user creates a project, adds an API key, and registers the agent or automation that will request approval.
project: vinac-production
agent: deployment-runner
environment: productionMeet agents where they take action.
Start with a small set of honest SDKs and adapters, then attach the local gateway for VINAC-FM, QR, and NFC proofs.
The primary product is controlled execution. Site-protection packages should stay secondary unless that product is intentionally revived.
One open plan until the workflow is proven
Every preview user should be able to protect an action, approve it, verify the receipt, and test gateway proof before pricing returns.
Open Preview
All useful platform features are available to every account while the approval workflow is hardened.
Before launch
Paid plans should come back only after usage, gateway reliability, receipts, and support workflows are proven.