Neura Local is available for Mac

Neura Relay controls the moment before agents act

Review and control important AI actions before they happen

Where Relay fits

A clear check happens before anything important changes

01

Understand the action

Relay reads what the agent is about to do before tools, data, or workflows change

02

Check permission and risk

Relay checks who is acting, what they can do, what evidence exists, and how risky the action is

03

Return a decision

Relay returns proceed, revise, review, or stop for that exact action

04

Keep an audit trail

Relay keeps a review record so systems and humans can see what happened and why

See how Relay reviews actions

Control costly agent actions before execution

Before money, data, production systems, messages, or workflows are affected

Commercial control point

Relay reviews agent actions before they affect money, data, production, messages, or workflows. The next system gets a clear decision and reason before anything changes.

Start with one protected action family. Map the review path, test stale, changed, and replayed actions, then see where agents create cost, risk, review load, and failed outcomes.

Decision records

Show who approved what, for which action, and why before work continues.

Traffic review

Find risky actions, review pressure, stale approvals, and repeated stops in one view.

Commerce checks

Review refunds, payouts, discounts, subscriptions, and purchases before money moves.

First guardrail

Turn one real workflow into a review step agents can use before execution.

Review a proposed action

Example 1 of 4

Rotating examples

Use arrows for all four examples. Playground runs full review.

Review result

Decision
Ready

Ready to proceed

Confidence
High

Recommended next step

The action may continue

Reason

The action has approved content, a bounded recipient, and no irreversible change

High confidence to proceed, score range 78–92

Add Relay before important actions

Send the proposed action to Relay and get a clear decision back

Integration path
  1. 01

    Describe the action

    Proposed action and actor

  2. 02

    Relay reviews it

    Relay checks permission, evidence, risk, and timing before anything happens

  3. 03

    Use the decision

    Continue, revise, review, or stop before work affects money, data, systems, or workflows

Resolve call
Example
TypeScript
const response = await fetch("/api/resolve", {  method: "POST",  headers: { "Content-Type": "application/json" },  body: JSON.stringify({ action_card: actionCard })});
const response = await fetch("/api/resolve", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ action_card: actionCard }) }); const { decision_receipt: receipt } = await response.json(); if (receipt.decision === "proceed") { execute(actionCard); } else { queueHumanReview(receipt); }

Before agents act

Send the proposed action in and get a clear decision back

A 63-second film introducing the path from proposed action to clear decision.

Open on YouTube

One path across Relay, Registry, and Protocol

Relay reviews the action, Registry identifies the agent, Protocol keeps the shared decision record

Questions before production

What Relay reviews, what agents can use, and how protected actions connect

Integration FAQ

Relay reviews important AI-agent actions before they change data, send messages, call tools, trigger workflows, or move money. It returns a decision another agent or system can use before anything changes.