Agent Observability Pack Generator — Guide

You cannot operate a payment agent in production without knowing when it is failing, slowing down, or approaching its limits. CloudWatch alarms, X-Ray tracing, and DLQ monitoring are the minimum observability requirements for any payment agent — but setting them up from scratch is time-consuming and error-prone.

The Agent Observability Pack Generator produces a complete observability configuration for your payment agent stack in seconds. Configure your stack name, environment, and feature toggles — the tool generates CloudWatch alarms, a dashboard, X-Ray tracing config, and DLQ monitoring for all 6 standard payment agent types.

The output satisfies PCI DSS Requirements 10.2, 10.3, and 10.5 — audit event logging, log protection, and log retention. File headers in the generated output reference these requirements explicitly so your QSA can map each resource to the control it satisfies.

1. Stack name: the identifier used as a prefix for all generated AWS resource names. Use lowercase and hyphens — e.g. payment-agents-prod. The tool auto-sanitises the name (converts uppercase to lowercase, removes invalid characters) to ensure it is a valid AWS resource name.
2. Environment: Production, Staging, or Dev. This controls log retention (365 / 90 / 14 days respectively), whether SNS alerts are generated, and the alarm evaluation periods. Production settings satisfy PCI DSS Req 10.5 log retention requirements.
3. X-Ray tracing: toggle on to include X-Ray tracing configuration in the output. Adds an X-Ray group resource and enables active tracing on Lambda functions. Recommended for Production — enables distributed trace analysis per payment execution.
4. DLQ monitoring: toggle on to add CloudWatch Alarm and dashboard widget for Dead Letter Queue message count. If the DLQ receives any messages, the alarm fires immediately — a non-zero DLQ count means at least one payment intent has failed all retries.

1. The 6 agent cards (Payment Router, Fraud Detector, Settlement Agent, Compliance Checker, Notification Agent, Webhook Processor) can each be toggled on or off.
2. Each selected agent adds its own set of alarms and a dashboard row to the output. Only select agents you are actually deploying — unused alarm resources add noise and cost.
3. The alarm preview panel (right side of each agent card) shows the colour-coded severity level (red = critical, amber = warning, purple = info) for each alarm before you generate.

1. Switch between Terraform HCL, CloudWatch Dashboard JSON, and CloudFormation YAML tabs. All output is real-time — no generate button needed.
2. Click Export to download the file as {stackName}-observability.{tf|json|yaml}.
3. For Terraform: place the .tf file in your infrastructure repo and run terraform apply. For Production, the SNS topic ARN will be output — subscribe your on-call tool (PagerDuty, Slack, email) to it.
4. For CloudWatch Dashboard JSON: go to CloudWatch → Dashboards → Create Dashboard → Source, paste the JSON, and save. The dashboard will populate with data once your agents are running.
5. For CloudFormation: run aws cloudformation deploy --template-file observability.yaml --stack-name {stackName}.