Skip to content

Trust and control

Powerful enough to write.
Careful enough to stop.

Netrix reaches live customer sites, agency credentials, and client data. The control model treats refusal, exact scope, verification, and recovery as product behavior—not implementation trivia.

Tenant-scoped access Encrypted credentials Explicit approval Reversible writes
NetrixDecision boundary
ReviewPerson decides
TargetExact identity
VerifyRead it back
RevertKnown prior value
Netrix at the center of four controls: human review, exact targeting, live verification, and recovery.

Safeguards by boundary

Controls belong where the risk enters.

A broad promise like “secure” says little. These are the concrete boundaries the application enforces.

01

Workspace boundary

Who can read or act

IDENTITY
Account data
Scoped through the current membership
Client access
Limited to explicitly granted businesses
Ownership
The last owner cannot be removed or demoted
02

Credential boundary

What a connector may use

SECRETS
At rest
Integration credentials are encrypted
At render
Stored secrets remain write-only in the UI
At use
Provider and account scope travel together
03

Network boundary

Where the server may reach

EGRESS
Addresses
Private and metadata targets are refused
Redirects
Every hop is validated again
Requests
Time and response-size budgets are bounded
04

Write boundary

What changes on a site

DELIVERY
Approval
Accepted work is distinct from a draft
Target
The page element carries an exact identity
Outcome
Applied, verified, failed, and reverted stay distinct

A live change as a transaction

The safe path has more than a Deploy button.

Each phase answers a different question, and a failure in one does not get mislabeled as success in another.

  1. 01

    Resolve

    Name the page and exact element the finding belongs to.

    Stops on ambiguity
  2. 02

    Authorize

    Check the acting membership, review state, connector, and site.

    Stops on scope
  3. 03

    Apply

    Write through the active connector and retain the prior value.

    Records the attempt
  4. 04

    Verify

    Fetch the live page and compare the action semantically.

    Separate outcome
  5. 05

    Recover

    Revert to the known prior value or release work after a crash.

    Durable trail

Product principles

Clear language is part of the control surface.

Netrix drafts.

People approve.

The platform does not describe a recommendation as a live change or promise hands-off autopilot.

Applied is not verified.

Both facts remain visible.

A connector accepting a write and a live page reflecting it are separate events.

Refusal is a feature.

Unclear work stops.

An ambiguous target, mismatched provider, exhausted budget, or unavailable dependency should produce an actionable refusal.

Recovery is designed.

Not left to a retry button.

Durable states, retained prior values, serialized reversals, and stale-work recovery make interruption part of the model.

Connector protocols

Trust is scoped to the path doing the work.

The WordPress plugin authenticates signed requests with replay defenses and its own permission gate. Cloudflare resources are adopted only with persisted ownership evidence. Legacy REST credentials stay tied to one verified site.

NETRIXPOST /applytimestamp · nonce · body digest · signature
SITE GATE
  • Signature
  • Freshness
  • Permission
LIVE SITEExact targetKnown prior value retained
A signed Netrix request passes through authentication, freshness, and permission checks before reaching an exact live-site target.

Plain answers

Trust questions

Does Netrix make live changes automatically?

No. Netrix finds issues and can draft remedies; a person reviews and accepts the work before a deployment is requested.

Can a client portal user change agency setup?

No. Client access is limited to the businesses granted to that user, and operational actions such as connector management, audit starts, AI spend, and deployment remain with agency staff.

What happens after a connector says a change succeeded?

Netrix records the write as applied, then independently reads the live page. Only that second check can mark the deployment verified.

Can a deployed change be undone?

Supported deployments retain the prior value and expose a serialized revert path, with the reversal recorded in the same deployment history.

See the system in context

Control is most useful when it stays out of the team's way.

Back to the product