Why should you care?

Without a shared pattern, CRM and ERP teams ship incompatible agent APIs that operations cannot monitor uniformly.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Agents/Architecture/AgentBase.cs
public abstract class AgentBase(IChatClient chat, IAgentTelemetry telemetry)
{
    protected async Task<T> ExecuteAsync<T>(string agentName, Func<Task<T>> work, CancellationToken ct)
    {
        using var activity = telemetry.Start(agentName);
        try { return await work(); }
        catch (Exception ex) { telemetry.Fail(agentName, ex); throw; }
    }
}

What happened?

• AgentBase wraps work with telemetry activities.
• Concrete CRM agents inherit consistent logging and error handling.

Practice next

1. Create AgentBase with telemetry hooks.
2. Inherit CrmCopilotAgent from AgentBase.
3. Define IAgentTelemetry with OpenTelemetry exporter.
4. Add optional IInputGuard on AgentBase before work runs.
5. Return AgentResult with status enum instead of raw payloads.

Remember

AgentBase standardizes telemetry and error paths. Agent architecture separates plan, tools, and format steps. Concrete agents stay thin inside ExecuteAsync.

Ops dashboard

SRE needs p95 latency per agent name across tenants.

Outcome: AgentBase telemetry feeds one Grafana board for all AgentNest agents.