Setting Up AI Agents with Agent Gateway

Agent Gateway is where AI meets automation in iDataWorkers. It lets you deploy AI agents that continuously monitor your data, detect patterns, and take action — without writing code. Think of agents as intelligent assistants that work around the clock on your behalf.

What You Will Learn

What AI agents are and how they differ from simple automations
How to create your first agent in Agent Gateway
How to configure triggers, actions, and guardrails
How to monitor agent activity and performance

Prerequisites

At least one data source connected via Data Sync
Basic familiarity with your data in Data Hub

Understanding AI Agents

Traditional automations follow rigid rules: "If X happens, do Y." AI agents are different — they reason about your data. An agent can:

Detect anomalies it was never explicitly told to look for
Correlate events across multiple data sources
Adapt its behavior based on historical patterns
Explain its reasoning in plain language

Step 1: Open Agent Gateway

Navigate to Agent Gateway from the sidebar. The dashboard shows your active agents, recent activity, and available agent templates.

Step 2: Create a New Agent

Click + New Agent. Choose from templates or start from scratch:

Anomaly Detector — flags unusual patterns in metrics
Data Quality Monitor — catches missing, duplicate, or malformed records
Trend Spotter — identifies emerging trends before they become obvious
Custom Agent — define your own objectives

Select Anomaly Detector for this walkthrough.

Step 3: Configure the Agent

Give your agent a name and description, then configure its parameters:

Name:        Revenue Anomaly Watcher
Data Source:  orders (from Data Hub)
Metric:       daily_revenue
Sensitivity:  Medium (2 standard deviations)
Check Every:  1 hour
Alert Via:    Email + Slack

Tip: Start with Medium sensitivity. You can fine-tune after the agent runs for a few days — too sensitive means too many alerts; too relaxed means missed anomalies.

Step 4: Set Guardrails

Guardrails prevent agents from overreacting. Configure:

Cooldown period — minimum time between alerts (e.g., 4 hours)
Confirmation threshold — require the anomaly to persist for N checks before alerting
Escalation path — who gets notified if the agent detects a critical issue

Step 5: Deploy and Monitor

Click Deploy Agent. The agent begins monitoring immediately. In the Activity Log, you can see:

Every check the agent performed
Anomalies detected (with explanations)
Actions taken (alerts sent, reports generated)
Agent confidence scores

Conclusion

Your first AI agent is now live and monitoring your data. Agent Gateway handles the orchestration — the agent reasons, you decide. Explore the other agent templates or build a custom agent for your specific use case.