AEGIS Sentinel

Events 0

Decisions 0

Agents 0

Alerts 0

Reviews 0

Workspace Control

Start with one controlled workflow

AEGIS is built for one clear use case: approval and audit for AI agents and autonomous workflows before they touch production systems.

01 Connect workflow

Create one agent key and paste the validation call before sensitive actions.

02 View risk

See the latest decision, risk level and recommended next step.

03 Approve actions

Approve once, deny once, or create recurring allow/block rules for repeated patterns.

04 Keep activity proof

Export customer-safe evidence of decisions, rules, approvals and expiry.

Customer Guide

How AEGIS controls agent actions

1 Agent asks first

The agent sends the intended action to AEGIS before touching a real system.

2 AEGIS validates

AEGIS checks context, behavior, policy signals and operational risk.

3 Human decides

Risky actions wait for approval, denial or a recurring customer rule.

4 Agent confirms permission

Before execution, the agent confirms the approval is still valid and still matches the action.

5 Activity proves control

Approvals, blocks, rules, confirmations and repeated matches remain visible in the activity trail.

Observer Monitor first. Build context without blocking workflows.

Guardian Review before execution. Require approval for controlled actions.

Decision Rules Allow or block recurring patterns without exposing internal logic.

False Positives If AEGIS holds a safe action, approve once or allow similar to reduce repeated reviews.

Action Review

Check one sensitive action

Guardian Observer Gateway

Use a safe sample to see the answer your agent receives before it touches a real system. Customer workspaces run review-first by default.

Workflow agent-alpha

Action alter_permissions

Target production-db

Advanced action details

Workflow Action type Actor Target Action

Optional context JSON

Decision Preview

Waiting

none learning none

Risk0

Confidence0

Signal0

Uncertainty0

No action validated yet.

Customer Portal

Billing

Current plan Not loaded

Status unknown

Period Not available

Usage

Current Month

Events0 / 0

Agents0 / 0

Retention0 days

Connectors0

Plan Management

Upgrade without losing data

Upgrades keep the same workspace, memory, activity history and agent keys. Only plan limits and billing change.

Invoices

Billing History

Onboarding

Next steps

1 Create API key

Generate one key for the first agent or automation.

2 Copy snippet

Paste the validation call before sensitive actions.

3 Send first event

Run one safe action and confirm the workspace is receiving decisions.

Support

Contact AEGIS

For onboarding, billing or security questions, contact AEGIS support.

aegis.sentinel.team@gmail.com

Guided Setup

Connect in three steps

Connect is the setup workspace: create an agent key, copy one validation snippet, then send a safe test action before the automation touches a real system. Keep the workflow in human-review-first mode until you trust the pattern.

Plan active

Create agent key

Create a key for the first agent or automation.

Run first validation

Run one safe validation to confirm the integration.

Agent Access

Generate agent key

Key name Actor scope

Create adds a key for a new agent or automation. Rotate is available only on the current key and replaces a lost secret. Older keys remain listed for compatibility without customer actions. Secrets are copied once, never displayed again. Billing period, limits and retention stay unchanged.

Existing keys

Step 2

Copy the validation call

Place one snippet immediately before the real action. If AEGIS asks for review, the agent pauses until a human decision is recorded.

cURL

Python

JavaScript

Runtime Rule

What the agent must do

ReviewSend intended action to AEGIS before execution.

PauseIf review is required, the agent waits and does not execute.

ConfirmAfter approval, confirm permission before running the original action.

ActivityApprovals, denials and recurring rules stay visible in the customer's activity view.

Customer deployments start in review-first mode. Managed execution is reserved until explicitly enabled.

Guide

Understand how AEGIS protects a workflow

This page explains the operating model. Use Connect only when you are ready to generate an agent key and paste the validation call into a real automation.

1 Agent asks first

The agent sends the intended action to AEGIS before touching a real system.

2 AEGIS reviews context

AEGIS compares the action with workspace history, policy signals, behavior and operational risk.

3 Human decides when needed

Risky actions wait for approval, denial, or a recurring customer rule.

4 Activity becomes evidence

Approvals, denials, expiry and rule changes remain visible without exposing internal engine logic.

Where to place AEGIS

Protect the actions that can hurt a business

Permission changes Admin access, security settings, database permissions and infrastructure changes.

External actions Email sends, CRM updates, API calls, webhook triggers and customer-facing automation.

Financial movement Refunds, payouts, credits, billing changes and any workflow that can move money.

Multi-agent handoffs Agent-to-agent work where the next step should be reviewed before propagation.

Automation Recipes

Where the validation call goes

n8n, Make or ZapierAdd an HTTP request step immediately before the sensitive step. Continue on ALLOW, pause on REQUIRE_REVIEW, stop on BLOCK.

Python or internal jobsWrap the function that changes data, permissions, infrastructure, money or external communication.

OpenAI agents and toolsCall AEGIS before tools that send messages, deploy code, update CRM, move money or touch customer data.

Approval resumeWhen a review is approved, confirm execution permission before the agent runs the original action.

Core Concepts

What the terms mean

ObserverMonitors events and builds behavior history without blocking the workflow.

GuardianValidates intended actions before execution and requires human review when risk is high.

ApprovalA human decision that lets one action continue or creates a recurring allow/block rule.

ActivityThe customer-safe record of decisions, approvals, denials, expiry and recurring rules.

Operating rule

Start review-first, then reduce noise safely

Approve onceUse when one action is safe but should not become automatic yet.

Deny onceUse when the action should not run this time.

Allow similarUse only after the pattern is trusted and should stop asking every time.

Block similarUse when a repeated pattern should always be stopped before execution.

Activity History

Every permission change in one place

1 Search actions

Find decisions by workflow, actor, decision type or mode.

2 Inspect outcome

See the decision, risk band, permission status and expiry without exposing private engine logic.

3 Export evidence

Download a customer-safe report for approvals, denials, rule changes and permission checks.

Activity Search

Find an action, approval or rule change

Source Actor Decision Mode

Inspector

No selection

Select a decision, agent, timeline item or activity event.

Timeline

What changed

Activity Export

Customer-safe evidence

Workspace Settings

Workspace defaults

Plan Loading plan...

Billing, invoices and upgrades are managed from Account.

Access Loading keys...

Agent API keys are generated from Connect and secrets are copied once, then hidden.

Execution Review-first

Guardian and Observer are available. Managed execution remains reserved.

Retention Loading retention...

Activity history is kept according to the current plan.

Notifications

Approval / action webhook

Use Review to notify a human channel. Use Action to continue a paused Make, Zapier or n8n workflow only after approval.

Name Kind Purpose URL

Safety Defaults

How this workspace runs

Human review firstRisky actions wait for approval before the agent can continue.

Customer rulesRecurring safe or unsafe patterns can be allowed or blocked by the customer.

Activity visibleApprovals, denials, expiry and recurring rules remain visible in Activity.

Billing & License

Plan Control

Plan Email Checkout ID

Billing not loaded yet.

Customer Onboarding

Account Setup

Org Admin Password Agent Key Actor

Complete a paid checkout before onboarding.

Internal Settings

Console Maintenance

Development/demo controls. In a customer workspace this stays restricted to internal operations.

Platform Intelligence

Platform Observability

Internal-only view of customers, trials, payments, abuse, usage, errors and growth. Customers do not see this layer.

Logins

Trials 0

Events 24h 0

Ready Agents 0

Payment Issues 0

Errors 0

Growth 7d 0

Abuse Flags 0

Customers & Plans

Payments, Trials & Errors

Learning Signals

Platform Console

Customers, Plans & Abuse

Internal-only. Customers cannot see this panel, global learning signals, payment signals or abuse controls.

Customers 0

Active Paid 0

Trials 0

Abuse Flags 0

Blocked 0

Events 0

Events 24h 0

Reviews 0

Errors 0

Failed Logins

Active Sessions 0

Operator Radar

What needs attention

Internal-only summary of trials, abuse, payment issues, errors and usage spikes.

Trial Control

Free Trial Policy

Internal-only. Final trial rules, expiry behavior and abuse signals are visible here before public selling.

Customers

Accounts

Open details, review usage by organization, block abusive trials or unblock trusted accounts.

Customer Detail

Usage & Recent Activity

Approval Workspace

Review risky actions without engine noise

Pending approvals Actions waiting for a human decision before the agent can continue.

Recent risk Latest controlled decisions and the recommended next step.

Known agents Workflows AEGIS has observed, with counts instead of private behavior models.

Decision rules Customer-created allow/block patterns for repeated safe or unsafe actions.

Approval Queue

Actions waiting for review

Risk History

Recent decisions

Monitored Workflows

Known agents

Execution Gateway

Gateway Runs

Universal Agent Layer

Integration Kit

Control Maturity

Learning Gate

Actor Mode Min History

Learning mode is active by default.

Safety Controls

Active guardrails

Customer Controls

Sentinel

Sign in to your workspace