Administering Agent Mesh

This section is the operator's day-two home — what you do once Agent Mesh is running. First-time setup (install commands, initial configuration, choosing a deploy target) lives in Install and Deploy; the pages here cover operating, securing, observing, and troubleshooting a live deployment.

Before you move your Solace Agent Mesh into production, we recommend that you review the sections that follow.

Production Checklist

• Production Readiness Checklist—We strongly recommend that you use this pre-production checklist prior to moving to production.

Security and Access

• Enabling Single Sign-On (SSO)—Configure OpenID Connect (OIDC) single sign-on: the provider catalog, callback URIs, and the login flow.

• Secure User-Delegated Tool Access—Let each user act on remote systems through their own OAuth credentials.

• Managing Secrets—Where secrets are read from, the substitution syntax that wires them into YAML, and how to rotate each one without downtime.

• Configuring TLS—TLS for the inbound listener, the event broker connection, and outbound OIDC, large language model (LLM), and Model Context Protocol (MCP) traffic.

• Managing Audit and Compliance—The audit event types, the closed schema, and how to ship audit records to your log aggregator.

Observability and Feedback

• Monitoring Your Agent Mesh—Operational logs, OpenTelemetry metrics, traceID correlation, and health endpoints.

• Collecting and Publishing User Feedback—Collect end-user feedback, publish it to the event mesh, and retrieve it through the API.

Data and Backups

• Managing Backups and Data Retention—What to back up, how log rotation and the data-retention sweep work, and what is delegated to the underlying storage.

The Broker Layer

• The Event Mesh Communication Layer—The broker layer Agent Mesh runs on: the runtime connection block, the namespace tenant fence, and the durable queues that appear on the broker.

Runbooks and Troubleshooting

• Operator Workflows—End-to-end walkthroughs: rotating an LLM credential without downtime, staging a pre-upgrade dry-run, and scaling the Agent-Workflow Executor.

• Troubleshooting a Running Deployment—Live failures organized by what broke, each with symptoms, diagnostics, resolution, and prevention.