A demo banking application for Azure SRE Agent. Failure scenarios are embedded into normal banking workflows — clicking buttons like "Generate Annual Statement" or "Run Fraud Detection" silently triggers realistic production issues (memory leaks, CPU spikes, HTTP 500s, etc.) that SRE Agent can detect, investigate, and mitigate.

graph TD
    subgraph ACA["Azure Container Apps"]
        subgraph WebApp["Contoso Bank Web App"]
            Razor["Razor Pages<br/>(Frontend)"] -->|requests| API["ASP.NET Core Web API<br/>Controllers + Services<br/>ChaosService (DI)"]
        end
    end

    API -->|"UseAzureMonitor()"| AppInsights["App Insights +<br/>Log Analytics"]
    API --> SQL[("Azure SQL<br/>Database")]
    AppInsights -->|"KQL queries"| Grafana["Managed Grafana<br/>(dashboard + MCP)"]

    AppInsights --> SRE["Azure SRE Agent<br/>(configured separately)"]
    Grafana --> SRE

• Azure Developer CLI (azd) v1.9+
• Azure CLI (az) v2.60+
• .NET 10 SDK
• Docker Desktop (for local container builds)
• An Azure subscription with permissions to create resources (Contributor + User Access Administrator)

# Clone the repo
git clone https://github.com/<org>/sre-agent-demo.git
cd sre-agent-demo

# Deploy everything (infra + app + monitoring + dashboards)
azd up

azd up provisions all Azure resources, builds and deploys the container, and runs the post-provision script which:

1. Grants managed identity access to SQL Database
2. Seeds the database with demo banking data
3. Configures Container Apps OpenTelemetry agent
4. Sets up Grafana data sources (Azure Monitor, Log Analytics)
5. Imports the Grafana dashboard
6. Outputs the Grafana MCP endpoint URL and SRE Agent setup instructions

# Build and run locally (uses EF Core InMemory database)
cd src/ContosoBank
dotnet run

# Or run with Docker
docker build -t contoso-bank .
docker run -p 8080:8080 contoso-bank

The app is available at http://localhost:8080 with these endpoints:

The grafana/ directory contains a Docker Compose setup that runs Grafana locally, connected directly to your Azure Monitor / Log Analytics workspace via managed identity — no Prometheus needed.

cd grafana
docker compose up -d
# Grafana is available at http://localhost:3000 (admin/admin)

The local Grafana instance is pre-provisioned with:

• Azure Monitor datasource — authenticates via managed identity (azureAuthType: msi)
• Contoso Bank dashboard — the same KQL-based dashboard deployed to Azure Managed Grafana

The demo includes 8 distinct failure scenarios, each triggered by a natural-looking banking action. Every scenario auto-recovers after 5 minutes so the demo can be repeated without manual cleanup.

📋 See DEMO.md for a detailed demo walkthrough with suggested flow and best starting scenarios.

Each chaos scenario produces distinct telemetry that SRE Agent can investigate:

graph TD
    subgraph DotNetApp[".NET 10 App (OpenTelemetry SDK)"]
        Logs["ILogger&lt;T&gt;<br/>(structured logs)"] --> OTel["OpenTelemetry SDK"]
        Metrics["System.Diagnostics.Metrics<br/>(custom 'ContosoBank' meter)"] --> OTel
        Traces["ActivitySource<br/>(custom spans)"] --> OTel

        OTel --> AzExporter["Azure Monitor<br/>Exporter"]
    end

    AzExporter --> AppInsights["App Insights +<br/>Log Analytics"]
    AppInsights -->|"KQL queries"| Grafana["Managed Grafana<br/>(dashboard)"]
    Grafana --> MCP["MCP endpoint"]

    AppInsights -->|"KQL queries"| SRE["SRE Agent"]
    MCP -->|"KQL queries"| SRE

Every Azure Managed Grafana instance exposes a built-in MCP endpoint at /api/azure-mcp (docs). After azd up, find your endpoint:

# Get the Grafana MCP endpoint from azd environment
azd env get-value GRAFANA_MCP_ENDPOINT
# Example: https://graf-<name>.cse.grafana.azure.com/api/azure-mcp

Use a Grafana service account token for a one-time setup that doesn't expire:

1. Open your Grafana instance (get the URL with azd env get-value GRAFANA_ENDPOINT)
2. Navigate to Administration → Service accounts
3. Create a new service account with Viewer role
4. Click Add token → generate a token (format: glsa_xxx)
5. Copy the token — you'll use it in the connector config below

Alternative: Use an Entra ID token for short-lived access (expires in ~1 hour):

az account get-access-token --resource ce34e7e5-485f-4d76-964f-b3d2b16d1e4f --query accessToken -o tsv

1. Go to sre.azure.com → Builder → Connectors
2. Click + Add connector → MCP Server
3. Enter the MCP URL from azd env get-value GRAFANA_MCP_ENDPOINT
4. Auth: Bearer token — paste the Grafana service account token (glsa_xxx)
5. Select tools → Select all
6. Save — status should show Connected

Through this MCP endpoint, SRE Agent gets access to tools like amgmcp_query_resource_log, amgmcp_insights_get_failures, amgmcp_prometheus_query, and more — enabling it to query App Insights, Log Analytics, and Azure Monitor metrics directly via Grafana.

├── azure.yaml                    # azd project manifest
├── DEMO.md                       # Demo cheat sheet & walkthrough
├── ContosoBank.slnx              # Solution file
├── grafana/
│   ├── docker-compose.yml        # Local Grafana (Azure Monitor via managed identity)
│   ├── dashboards/
│   │   └── contoso-bank.json     # Grafana dashboard (KQL panels)
│   └── provisioning/
│       ├── dashboards/
│       │   └── default.yaml      # Dashboard provisioning config
│       └── datasources/
│           └── azure-monitor.yaml # Azure Monitor datasource (MSI auth)
├── infra/
│   ├── main.bicep                # Main orchestration (subscription-scoped)
│   ├── main.bicepparam           # Parameter defaults
│   └── modules/                  # 8 Bicep modules (see Deployed Resources)
├── scripts/
│   ├── post-provision.sh         # Post-azd setup (SQL, Grafana, OTel)
│   ├── seed-data.sql             # Demo banking data (5 accounts, 112 txns)
│   └── validate-dashboard.sh     # Dashboard JSON validation
├── src/
│   └── ContosoBank/
│       ├── Controllers/          # API controllers (5)
│       ├── Data/                 # EF Core DbContext + seed data
│       ├── Interceptors/         # ChaosDbInterceptor (DB failure injection)
│       ├── Metrics/              # BankMetrics (custom OTel meters)
│       ├── Models/               # Account, Transaction, Transfer
│       ├── Pages/                # Razor Pages (Dashboard, Accounts, etc.)
│       ├── Services/             # Business logic + ChaosService
│       ├── Dockerfile            # Multi-stage .NET 10 build
│       └── Program.cs            # App config (OTel, health checks, DI)
├── tests/
│   └── ContosoBank.Tests/        # Unit + integration tests
└── specs/
    └── 01-spec.md                # Full specification

dotnet test

Tests use WebApplicationFactory with EF Core InMemory provider. No external dependencies required.

# Remove all Azure resources
azd down

# Remove with force (skip confirmation)
azd down --force --purge

This deletes the resource group and all resources within it, including the SQL database, Container App, Grafana instance, and all monitoring data.

This project is for demonstration purposes. See LICENSE for details.